yhz.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import mido\n",
    "from mido import MidiFile"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "file = MidiFile('data/Midi #8.mid')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "message = mido.Message('program_change', program=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Populating the interactive namespace from numpy and matplotlib\n"
     ]
    }
   ],
   "source": [
    "%pylab inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "notes = []\n",
    "for message in file:\n",
    "    try:\n",
    "        notes.append(message.note)\n",
    "    except AttributeError:\n",
    "        pass"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x106e7fba8>]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD8CAYAAABn919SAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAFrJJREFUeJzt3X+wXGV9x/H3JwlBScuEhIWmxBLiMKB1SoA1E0qHUbAK\nyADOaIcOtdTiRDtosXQGwviPdsqMHbWoMxQbQUgtihihMnRKzQQyHf8I9l4JMfKjxEAgEMkagz9n\nSML99o89Fy/J3rtnb/bsefbZz2vmzu6ee/Y+33Oes9+cPOd8n1VEYGZmw29O3QGYmVl/OKGbmWXC\nCd3MLBNO6GZmmXBCNzPLhBO6mVkmnNDNzDLhhG5mlgkndDOzTMwbZGPHH398LFu2bJBNmpkNvfHx\n8Z9GRKPbegNN6MuWLWNsbGyQTZqZDT1JO8us5yEXM7NMOKGbmWXCCd3MLBNO6GZmmXBCNzPLRNeE\nLuk0SVum/PxC0ickLZK0QdLTxeNxgwjYzMw663rbYkQ8BawAkDQXeAG4D1gDbIyIz0haU7y+ocJY\n+2Z85z4279jLccfMZ99v9rNq+WLOPrn6f4/qareMydhWLV8M8NrzquMru0+879Jpt6wc+iz1fXyo\nXu9DvwD4cUTslHQZ8I5i+TpgE0OQ0Md37uPK2zbzyoEJApgjmD9vDnd9eFWlnVRXu73Etv/gBPPm\nCCQOvjpReXxl94n3XTrt9hrfMPdZ6vu4k17H0K8AvlE8PzEidgMUjyd0eoOk1ZLGJI21Wq3ZR9on\nm3fsZf/B9kEGMBFw4OAEm3fszbLdMiZjmwg48GpwYPJ5xfGV3Sfed+m022t8w9xnqe/jTkondEnz\ngUuBb/XSQESsjYhmRDQbja6Vq5VbtXwx8+fNeW3D5wiOmjfntf9S5dZuGZOxzRUcNVccNfm84vjK\n7hPvu3Ta7TW+Ye6z1PdxJ4qI7msBxRDLNRHx7uL1U8A7ImK3pCXApog4baa/0Ww2I4XSf4+hH67u\ncWCPoQ9Pu2Xl0Gep7GNJ4xHR7LpeDwn9buC/I+KO4vVngb1TLoouiojrZ/obdSR0J+/DpX4wp5zk\n694ndSeW6aTcZ2Wl/Lkom9BLXRSVdAzwp8BHpiz+DHCPpKuB54APzCbQKvkC6PSxpXpBKOULpXXv\nk1QvzqXcZ2Wl/rkoq9QYekT8JiIWR8TPpyzbGxEXRMSpxePPqgtzdnwB9HCpXxBK+UJp3fsk1Ytz\nKfdZWal/LsrKulLUF0APl/oFoZQvlNa9T1K9OJdyn5WV+ueirNJj6P3gMXSPFfYSX4rjsXXvE4+h\nVyflz0XfL4r2Qyp3uZiZDZOyCT3rIRczs1HihG5mlgkndDOzTDihm5llotfZFpPR6Urz1Cvnk8v6\nffW501X6Tu33+4r3oe12a3+mZXXdldEp9qr3Uy9t1NG3qR/HZZdV1Y8z3WVS17HdLd46YxnKhN6p\nWuvAwd9Wn1VVwdWp0q1T+/2uGju0XcGM7Q9yn8wUb6f+mRp71fuplzbq6NvUj+Oyy6rqx/0Hp6/U\n7BR7XRWbKVWPDuWQS6dqrddVn1VUwdWx0q1T+32uGju03a7tD3CfzBRvp1gGuZ96aaOOvk3+OC67\nrKJ+nKlSs65ju2u8NccylAm9U7XW66rPKqrg6ljp1qn9PleNHdquurU/wH0yU7yd+ud1sVe8n3pp\no46+Tf44Lruson6cqVKzrmO7a7w1xzK0hUWpjz16DN1j6L20mepxXHaZx9CrHUN3paiZWSZcKWpm\nNmKc0M3MMuGEbmaWiVIJXdJCSeslPSnpCUnnSPqUpBckbSl+Lq46WDMzm17ZwqIvAg9GxPslzQeO\nAd4D3BwRn6ssOjMzK61rQpd0LHAe8FcAEbEf2C9ppreZmdmAlRlyWQ60gDskPSrpNkkLit99TNJW\nSV+VlMbXjpiZjagyCX0ecBZwa0ScCfwaWAPcCrwZWAHsBj7f6c2SVksakzTWarX6E7WZmR2mTELf\nBeyKiEeK1+uBsyLipYh4NSImgK8AKzu9OSLWRkQzIpqNRqM/UZuZ2WG6JvSI+AnwvKTTikUXAI9L\nWjJltfcB2yqIz8zMSip7l8vHgbuKO1x2AB8CviRpBe0J9J4FPlJJhGZmVkqphB4RW4BD5xH4YP/D\nMTOz2XKlqJlZJpzQzcwy4YRuZpYJJ3Qzs0w4oZuZZcIJ3cwsE07oZmaZcEI3M8uEE7qZWSac0M3M\nMuGEbmaWCSd0M7NMOKGbmWXCCd3MLBNO6GZmmXBCNzPLRKmELmmhpPWSnpT0hKRzJC2StEHS08Xj\ncVUHa2Zm0yt7hv5F4MGIOB04A3gCWANsjIhTgY3FazMzq0nXhC7pWOA84HaAiNgfES8DlwHritXW\nAZdXFaSZmXVX5gx9OdAC7pD0qKTbJC0AToyI3QDF4wkVxmlmZl2USejzgLOAWyPiTODX9DC8Imm1\npDFJY61Wa5ZhmplZN2US+i5gV0Q8UrxeTzvBvyRpCUDxuKfTmyNibUQ0I6LZaDT6EbOZmXXQNaFH\nxE+A5yWdViy6AHgcuB+4qlh2FfCdSiI0M7NS5pVc7+PAXZLmAzuAD9H+x+AeSVcDzwEfqCZEMzMr\no1RCj4gtQLPDry7obzhmZjZbrhQ1M8uEE7qZWSac0M3MMuGEbmaWCSd0M7NMOKGbmWXCCd3MLBNO\n6GZmmXBCNzPLhBO6mVkmnNDNzDLhhG5mlgkndDOzTDihm5llwgndzCwTTuhmZplwQjczy0SphC7p\nWUk/lLRF0lix7FOSXiiWbZF0cbWhmpnZTMp+pyjAOyPip4csuzkiPtfPgMzMbHY85GJmlomyCT2A\n70oal7R6yvKPSdoq6auSjuv0RkmrJY1JGmu1WkccsJmZdVY2oZ8bEWcBFwHXSDoPuBV4M7AC2A18\nvtMbI2JtRDQjotloNPoRs5mZdVAqoUfEi8XjHuA+YGVEvBQRr0bEBPAVYGV1YZqZWTddE7qkBZJ+\nd/I58G5gm6QlU1Z7H7CtmhDNzKyMMne5nAjcJ2ly/a9HxIOSviZpBe3x9WeBj1QWpZmZddU1oUfE\nDuCMDss/WElEZmY2K75t0cwsE07oZmaZcEI3GyHjO/dxy8PbGd+5r+5QDlNVbHVtcx3t9lL6b2ZD\nbHznPq68bTP7D04wf94c7vrwKs4+uWM94MBVFVtd21xXuz5DNxsRm3fsZf/BCSYCDhycYPOOvXWH\n9JqqYqtrm+tq1wndbESsWr6Y+fPmMFdw1Lw5rFq+uO6QXlNVbHVtc13tKiIG0hBAs9mMsbGxgbVn\nZq83vnMfm3fsZdXyxckMt0yqKra6trmf7Uoaj4hm1/Wc0M3M0lY2oXvIxcwsE07oZmaZcEI3M8uE\nE7qZWSac0M3MMuGEbmaWCSd0M7NMOKGbmWWi1ORckp4Ffgm8ChyMiKakRcA3gWW0v7HozyIivSnc\nzMxGRC9n6O+MiBVTqpXWABsj4lRgY/HazMxqciRDLpcB64rn64DLjzwcG0Ypz7EN6cdn0+u170a9\nr8vOhx7AdyUF8K8RsRY4MSJ2A0TEbkknVBWkpSvlObYh/fhser32nfu6/Bn6uRFxFnARcI2k88o2\nIGm1pDFJY61Wa1ZBWrpSnmMb0o/Pptdr37mvSyb0iHixeNwD3AesBF6StASgeNwzzXvXRkQzIpqN\nRqM/UVsyUp5jG9KPz6bXa9+5r0tMnytpATAnIn5ZPN8A/ANwAbA3Ij4jaQ2wKCKun+lvefrcPKU8\nxzakH59Nr9e+y7Wv+zYfuqTltM/KoT3m/vWIuEnSYuAe4A+A54APRMTPZvpbTuhmZr0rm9C7XhSN\niB3AGR2W76V9lm5mZglwpaiZWSac0M3MMuGEbmaWCSf0hKVe9ZZ6fGajpmylqA1Y6lVvqcdnNop8\nhp6o1KveUo/PbBQ5oScq9aq31OMzG0VdC4v6yYVFvUm96i31+Mxy0bfCIqvP2Scfl3SiTD0+s1Hj\nIRczs0w4oZuZZcIJ3cwsE07oZmaZcEKfwSAqIV1tOTp8PFnVfJfLNAZRCelqy9Hh48kGwWfo0xhE\nJaSrLUeHjycbhNIJXdJcSY9KeqB4faekZyRtKX5WVBfm4A2iEtLVlqPDx5MNQulKUUnXAU3g2Ii4\nRNKdwAMRsb5sY8NWKTqISkhXW44OH082W337TtHijy0F1gE3AdeNSkI3M0tB2YRedsjlC8D1wMQh\ny2+StFXSzZKO7jVIMzPrn64JXdIlwJ6IGD/kVzcCpwNvBxYBN0zz/tWSxiSNtVqtI43XzMymUeYM\n/VzgUknPAncD50v694jYHW2vAHcAKzu9OSLWRkQzIpqNRqNvgZuZ2et1TegRcWNELI2IZcAVwEMR\n8ReSlgBIEnA5sK3SSM3MbEZHUlh0l6QGIGAL8NH+hGRmZrPRU0KPiE3ApuL5+RXEY2Zms+RKUTOz\nTDihm5llwgndzCwTQ5HQPe2o9VvV/e3jyeqQ/PS5nnbU+q3q/vbxZHVJ/gzd045av1Xd3z6erC7J\nJ3RPO2r9VnV/+3iyupSePrcfZjvboqcdtX6rur99PFk/9XX63H7x9LlmZr3r9/S5ZmaWOCd0M7NM\nOKGbmWXCCd3MLBNO6GaZS71qdZDx1bUvBtVu8pWiZjZ7qVetDjK+uvbFINv1GbpZxlKvWh1kfHXt\ni0G264RulrHUq1YHGV9d+2KQ7ZYuLJI0FxgDXoiISySdQvtLoxcBPwA+GBH7Z/obLiwyG7zUq1YH\nGV9d++JI2+17paik64AmcGyR0O8B7o2IuyV9GXgsIm6d6W84oZuZ9a6vlaKSlgLvBW4rXgs4H1hf\nrLIOuHx2oZqZWT+UHUP/AnA9MFG8Xgy8HBEHi9e7gJM6vVHSakljksZardYRBWtmZtPrmtAlXQLs\niYjxqYs7rNpx7CYi1kZEMyKajUZjlmGamVk3Ze5DPxe4VNLFwBuAY2mfsS+UNK84S18KvFhdmGZm\n1k3XM/SIuDEilkbEMuAK4KGIuBJ4GHh/sdpVwHcqi9LMzLo6kvvQbwCuk7Sd9pj67f0JyczMZqOn\n0v+I2ARsKp7vAFb2PyQzM5sNV4qamWXCCd3MLBNO6NZXKU/VmnJs1l2v/TeK/e3pc61vUp6qNeXY\nrLte+29U+9tn6NY3KU/VmnJs1l2v/Teq/e2Ebn2T8lStKcdm3fXaf6Pa36VnW+wHz7aYv5Snak05\nNuuu1/7Lqb/7Pn1uPzihm5n1rq/T55qZWfqc0M3MMuGEbmaWCSd0M7NMOKGbmWXCCT1BKZcspxyb\n2ahz6X9iUi5ZTjk2M/MZenJSLllOOTYzK/cl0W+Q9H1Jj0n6kaRPF8vvlPSMpC3Fz4rqw81fyiXL\nKcdmZiUqRSUJWBARv5J0FPA94Frgo8ADEbG+bGOuFC0n5ZLllGMzy1XZStGuY+jRzvi/Kl4eVfwM\nbr6AEXT2ycclmyxTjs1s1JUaQ5c0V9IWYA+wISIeKX51k6Stkm6WdHRlUZqZWVelEnpEvBoRK4Cl\nwEpJbwNuBE4H3g4sAm7o9F5JqyWNSRprtVp9CtvMzA7V010uEfEysAm4MCJ2R9srwB3AymneszYi\nmhHRbDQaRxywmZl1VuYul4akhcXzNwLvAp6UtKRYJuByYFuVgZqZ2czKFBYtAdZJmkv7H4B7IuIB\nSQ9JagACttC+6yVLg7izw3ePjBYfU1aFMne5bAXO7LD8/EoiSswgqiNdgTlafExZVVwp2sUgqiNd\ngTlafExZVZzQuxhEdaQrMEeLjymrir9TtASPd1q/+ZiyXvhLos3MMuEviTYzGzFO6GZmmXBCNzPL\nhBO6mVkmnNDNzDLhhG5mlgkndDOzTDihm5llwgndzCwTQ5fQx3fu45aHtzO+c99Q/n1LyyD628eU\nDUqZ+dCTUfWUoJ5ydLR4GlvLzVCdoVc9JainHB0tnsbWclPmK+jeIOn7kh6T9CNJny6WnyLpEUlP\nS/qmpPlVB1v1lKCecnS0eBpby03X2RaL7wxdEBG/knQU8D3gWuA64N6IuFvSl4HHIuLWmf5WP2Zb\nrHpKUE85Olo8ja0Ng0qmz5V0DO2E/jfAfwK/FxEHJZ0DfCoi3jPT+z19rplZ7/o6fa6kuZK2AHuA\nDcCPgZcj4mCxyi7gpNkGa2ZmR65UQo+IVyNiBbAUWAm8pdNqnd4rabWkMUljrVZr9pGamdmMerrL\nJSJeBjYBq4CFkiZve1wKvDjNe9ZGRDMimo1G40hiNTOzGZS5y6UhaWHx/I3Au4AngIeB9xerXQV8\np6ogzcysuzKFRUuAdZLm0v4H4J6IeEDS48Ddkv4ReBS4vcI4zcysi64JPSK2Amd2WL6D9ni6mZkl\noKfbFo+4MakF7Jzl248HftrHcOqSw3bksA2Qx3bksA2Qx3ZUuQ0nR0TXi5ADTehHQtJYmfswU5fD\nduSwDZDHduSwDZDHdqSwDUM1l4uZmU3PCd3MLBPDlNDX1h1An+SwHTlsA+SxHTlsA+SxHbVvw9CM\noZuZ2cyG6QzdzMxmMBQJXdKFkp6StF3SmrrjKUPSmyQ9LOmJYh75a4vliyRtKOaR3yAp+flUi8nZ\nHpX0QPF64HPhHylJCyWtl/Rk0SfnDGlf/F1xPG2T9I3i+wqS7g9JX5W0R9K2Kcs67nu1fan4rG+V\ndFZ9kb/eNNvx2eKY2irpvsmq+uJ3Nxbb8ZSkGWei7ZfkE3pRoXoLcBHwVuDPJb213qhKOQj8fUS8\nhfbcN9cUca8BNkbEqcDG4nXqrqU93cOkfwJuLrZhH3B1LVH15ovAgxFxOnAG7e0Zqr6QdBLwt0Az\nIt4GzAWuIP3+uBO48JBl0+37i4BTi5/VwIzfsTBgd3L4dmwA3hYRfwT8H3AjQPFZvwL4w+I9/1Lk\nskoln9BpV6Nuj4gdEbEfuBu4rOaYuoqI3RHxg+L5L2knkJNox76uWG0dcHk9EZYjaSnwXuC24rWA\n84H1xSrDsA3HAudRTE8REfuLieaGqi8K84A3FhPjHQPsJvH+iIj/AX52yOLp9v1lwL9F22bakwAu\nGUykM+u0HRHx3SnTiG+mPVEhtLfj7oh4JSKeAbYzgMr6YUjoJwHPT3k9dHOvS1pGe/qER4ATI2I3\ntJM+cEJ9kZXyBeB6YKJ4vZjhmwt/OdAC7iiGjm6TtIAh64uIeAH4HPAc7UT+c2Cc4esPmH7fD/Pn\n/a+B/yqe17Idw5DQ1WHZ0NyaI+l3gG8Dn4iIX9QdTy8kXQLsiYjxqYs7rJp6f8wDzgJujYgzgV+T\n+PBKJ8U482XAKcDvAwtoD1EcKvX+mMkwHl9I+iTtYda7Jhd1WK3y7RiGhL4LeNOU19POvZ6a4jtY\nvw3cFRH3FotfmvwvZPG4p674SjgXuFTSs7SHus6nfcZeai78hOwCdkXEI8Xr9bQT/DD1BbSnrn4m\nIloRcQC4F/hjhq8/YPp9P3Sfd0lXAZcAV8Zv7wOvZTuGIaH/L3BqcSV/Pu0LDffXHFNXxVjz7cAT\nEfHPU351P+354yHxeeQj4saIWBoRy2jv94ci4kqGbC78iPgJ8Lyk04pFFwCPM0R9UXgOWCXpmOL4\nmtyOoeqPwnT7/n7gL4u7XVYBP58cmkmRpAuBG4BLI+I3U351P3CFpKMlnUL7Iu/3Kw8oIpL/AS6m\nfQX5x8An646nZMx/Qvu/WFuBLcXPxbTHoDcCTxePi+qOteT2vAN4oHi+vDg4twPfAo6uO74S8a8A\nxor++A/guGHsC+DTwJPANuBrwNGp9wfwDdpj/gdon7lePd2+pz1UcUvxWf8h7Tt6at+GGbZjO+2x\n8snP+JenrP/JYjueAi4aRIyuFDUzy8QwDLmYmVkJTuhmZplwQjczy4QTuplZJpzQzcwy4YRuZpYJ\nJ3Qzs0w4oZuZZeL/AUNh2F8AsAI/AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1056f2ac8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(notes,'.')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from keras.layers import LSTM, Dense, Dropout, Flatten, TimeDistributed\n",
    "from keras.layers import Input, Embedding\n",
    "from keras.models import Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "13\n",
      "104\n",
      "(104, 13)\n"
     ]
    }
   ],
   "source": [
    "sequence_length = 20\n",
    "feature_dim=1\n",
    "# get all pitch names\n",
    "pitchnames = sorted(set(item for item in notes))\n",
    "n_catagory=len(pitchnames)\n",
    "print (n_catagory)\n",
    "# create a dictionary to map pitches to integers\n",
    "note_to_int = dict((note, number) for number, note in enumerate(pitchnames))\n",
    "network_input = []\n",
    "network_output = []\n",
    "# create input sequences and the corresponding outputs\n",
    "for i in range(0, len(notes) - sequence_length, 1):\n",
    "    sequence_in =  notes[i:i + sequence_length]\n",
    "    sequence_out = notes[i + sequence_length]\n",
    "    network_input.append([note_to_int[char] for char in sequence_in])\n",
    "    network_output.append(note_to_int[sequence_out])\n",
    "n_patterns = len(network_input)\n",
    "print (n_patterns)\n",
    "# reshape the input into a format compatible with LSTM layers\n",
    "network_input = numpy.reshape(network_input, (n_patterns, sequence_length,feature_dim))\n",
    "network_output = keras.utils.to_categorical(network_output,num_classes=n_catagory)\n",
    "print (network_output.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n",
      "104/104 [==============================] - 1s 13ms/step - loss: 0.2545\n",
      "Epoch 2/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.2271\n",
      "Epoch 3/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.2233\n",
      "Epoch 4/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.2218\n",
      "Epoch 5/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.2200A: 0s - loss: 0.213\n",
      "Epoch 6/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.2212\n",
      "Epoch 7/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.2177\n",
      "Epoch 8/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.2194\n",
      "Epoch 9/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.2142\n",
      "Epoch 10/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.2131A: 0s - loss: 0.217\n",
      "Epoch 11/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.2107\n",
      "Epoch 12/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.2062\n",
      "Epoch 13/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.2075\n",
      "Epoch 14/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.2037\n",
      "Epoch 15/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.2005\n",
      "Epoch 16/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1990\n",
      "Epoch 17/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1914\n",
      "Epoch 18/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1948\n",
      "Epoch 19/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1873\n",
      "Epoch 20/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1878\n",
      "Epoch 21/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1901\n",
      "Epoch 22/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1827\n",
      "Epoch 23/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1855\n",
      "Epoch 24/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1812\n",
      "Epoch 25/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1828\n",
      "Epoch 26/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1782\n",
      "Epoch 27/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1765\n",
      "Epoch 28/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1781\n",
      "Epoch 29/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1745\n",
      "Epoch 30/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1710\n",
      "Epoch 31/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1687\n",
      "Epoch 32/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1692\n",
      "Epoch 33/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1673\n",
      "Epoch 34/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1692\n",
      "Epoch 35/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1703\n",
      "Epoch 36/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1688\n",
      "Epoch 37/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1672\n",
      "Epoch 38/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1637\n",
      "Epoch 39/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1636\n",
      "Epoch 40/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1624\n",
      "Epoch 41/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1633\n",
      "Epoch 42/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1643\n",
      "Epoch 43/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1590\n",
      "Epoch 44/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1617\n",
      "Epoch 45/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1561\n",
      "Epoch 46/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1648\n",
      "Epoch 47/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1544\n",
      "Epoch 48/200\n",
      "104/104 [==============================] - 0s 2ms/step - loss: 0.1521\n",
      "Epoch 49/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1568\n",
      "Epoch 50/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1496\n",
      "Epoch 51/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1520\n",
      "Epoch 52/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1532\n",
      "Epoch 53/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1546\n",
      "Epoch 54/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1519\n",
      "Epoch 55/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1539\n",
      "Epoch 56/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1544\n",
      "Epoch 57/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1452\n",
      "Epoch 58/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1434\n",
      "Epoch 59/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1520\n",
      "Epoch 60/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1432\n",
      "Epoch 61/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1483\n",
      "Epoch 62/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1368\n",
      "Epoch 63/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1475\n",
      "Epoch 64/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1383\n",
      "Epoch 65/200\n",
      "104/104 [==============================] - 0s 2ms/step - loss: 0.1322\n",
      "Epoch 66/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1423\n",
      "Epoch 67/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1470\n",
      "Epoch 68/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1331\n",
      "Epoch 69/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1386\n",
      "Epoch 70/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1290\n",
      "Epoch 71/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1416\n",
      "Epoch 72/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1356\n",
      "Epoch 73/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1364\n",
      "Epoch 74/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1299\n",
      "Epoch 75/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1317\n",
      "Epoch 76/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1277\n",
      "Epoch 77/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1251\n",
      "Epoch 78/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1256\n",
      "Epoch 79/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1180\n",
      "Epoch 80/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1382\n",
      "Epoch 81/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1320\n",
      "Epoch 82/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1213\n",
      "Epoch 83/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1264\n",
      "Epoch 84/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1243\n",
      "Epoch 85/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1264\n",
      "Epoch 86/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1285\n",
      "Epoch 87/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1153\n",
      "Epoch 88/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1290\n",
      "Epoch 89/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1147\n",
      "Epoch 90/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1375\n",
      "Epoch 91/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1233\n",
      "Epoch 92/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1218\n",
      "Epoch 93/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1144\n",
      "Epoch 94/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1115\n",
      "Epoch 95/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1175\n",
      "Epoch 96/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1153\n",
      "Epoch 97/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1131\n",
      "Epoch 98/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1130\n",
      "Epoch 99/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1133\n",
      "Epoch 100/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1025\n",
      "Epoch 101/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1243\n",
      "Epoch 102/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1084\n",
      "Epoch 103/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1162\n",
      "Epoch 104/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0993\n",
      "Epoch 105/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1040\n",
      "Epoch 106/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1049\n",
      "Epoch 107/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1064\n",
      "Epoch 108/200\n",
      "104/104 [==============================] - 0s 992us/step - loss: 0.1013\n",
      "Epoch 109/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1074\n",
      "Epoch 110/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0973\n",
      "Epoch 111/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1108\n",
      "Epoch 112/200\n",
      "104/104 [==============================] - 0s 995us/step - loss: 0.1001\n",
      "Epoch 113/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0993\n",
      "Epoch 114/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0939\n",
      "Epoch 115/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1089\n",
      "Epoch 116/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1017\n",
      "Epoch 117/200\n",
      "104/104 [==============================] - 0s 997us/step - loss: 0.0835\n",
      "Epoch 118/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0933\n",
      "Epoch 119/200\n",
      "104/104 [==============================] - 0s 981us/step - loss: 0.0830\n",
      "Epoch 120/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1045\n",
      "Epoch 121/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1096\n",
      "Epoch 122/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0943\n",
      "Epoch 123/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.1020\n",
      "Epoch 124/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0877\n",
      "Epoch 125/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0868\n",
      "Epoch 126/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0996\n",
      "Epoch 127/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0862\n",
      "Epoch 128/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0861\n",
      "Epoch 129/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0835\n",
      "Epoch 130/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0873\n",
      "Epoch 131/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0954\n",
      "Epoch 132/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0908\n",
      "Epoch 133/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0839\n",
      "Epoch 134/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0726\n",
      "Epoch 135/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0946\n",
      "Epoch 136/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0798\n",
      "Epoch 137/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0884\n",
      "Epoch 138/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0899\n",
      "Epoch 139/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0783\n",
      "Epoch 140/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0752\n",
      "Epoch 141/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0747\n",
      "Epoch 142/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0721\n",
      "Epoch 143/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0887\n",
      "Epoch 144/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0655\n",
      "Epoch 145/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0878\n",
      "Epoch 146/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0777\n",
      "Epoch 147/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0706\n",
      "Epoch 148/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0686\n",
      "Epoch 149/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0702\n",
      "Epoch 150/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0716\n",
      "Epoch 151/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0791\n",
      "Epoch 152/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0796\n",
      "Epoch 153/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0670\n",
      "Epoch 154/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0604\n",
      "Epoch 155/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0590\n",
      "Epoch 156/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0766\n",
      "Epoch 157/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0675\n",
      "Epoch 158/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0533\n",
      "Epoch 159/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0720\n",
      "Epoch 160/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0598\n",
      "Epoch 161/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0719\n",
      "Epoch 162/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0575\n",
      "Epoch 163/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0735\n",
      "Epoch 164/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0645\n",
      "Epoch 165/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0536\n",
      "Epoch 166/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0509\n",
      "Epoch 167/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0523\n",
      "Epoch 168/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0692\n",
      "Epoch 169/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0697\n",
      "Epoch 170/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0567\n",
      "Epoch 171/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0474\n",
      "Epoch 172/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0510\n",
      "Epoch 173/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0529\n",
      "Epoch 174/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0508\n",
      "Epoch 175/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0444\n",
      "Epoch 176/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0669\n",
      "Epoch 177/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0654\n",
      "Epoch 178/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0434\n",
      "Epoch 179/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0544\n",
      "Epoch 180/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0432\n",
      "Epoch 181/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0443\n",
      "Epoch 182/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0429\n",
      "Epoch 183/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0672\n",
      "Epoch 184/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0527\n",
      "Epoch 185/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0435\n",
      "Epoch 186/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0660\n",
      "Epoch 187/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0471\n",
      "Epoch 188/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0506\n",
      "Epoch 189/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0419\n",
      "Epoch 190/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0381\n",
      "Epoch 191/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0457\n",
      "Epoch 192/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0617\n",
      "Epoch 193/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0608\n",
      "Epoch 194/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0349\n",
      "Epoch 195/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0323\n",
      "Epoch 196/200\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0332\n",
      "Epoch 197/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0436\n",
      "Epoch 198/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0404\n",
      "Epoch 199/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0301\n",
      "Epoch 200/200\n",
      "104/104 [==============================] - 0s 1ms/step - loss: 0.0383\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x181e34bac8>"
      ]
     },
     "execution_count": 69,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "main_input = Input(shape=(sequence_length,feature_dim), name='main_input')\n",
    "#x = Embedding(output_dim=512, input_dim=(n_catagory,2), input_length=100)(main_input)\n",
    "lstm_out1 = LSTM(128,return_sequences=True)(main_input)\n",
    "lstm_out2 = LSTM(64)(lstm_out1)\n",
    "main_output = Dense(13, activation='softmax', name='main_output')(lstm_out2)\n",
    "model = Model(inputs=[main_input], outputs=[main_output])\n",
    "model.compile(optimizer='rmsprop',\\\n",
    "              loss={'main_output': 'binary_crossentropy'},\\\n",
    "              loss_weights={'main_output': 1.})\n",
    "\n",
    "# And trained it via:\n",
    "model.fit({'main_input': network_input},\\\n",
    "          {'main_output': network_output},\\\n",
    "          epochs=200, batch_size=32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "testing.mid\r\n"
     ]
    }
   ],
   "source": [
    "!ls featureToMidi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[12 12 12  1  1  1 10 10 10 10 10 10 10 10 11 11 11 12 12 11  2  7  8  3\n",
      " 10 10 10 10 12 12 11 11 12 12 12 11 10 10  8  3  5  0 10 10 10 10 11 11\n",
      " 12 12  6  6  6  1 10 10 10 10 11 11 12 12 12  1  1  1 10 10 10 10 10 10\n",
      " 10 10 11 11 11 12 11 11  4  9 10  8  3 10 10 10  8  3  9  4 12 12 11 11\n",
      " 12 12 11 11 10 10  9  9]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "int_to_note = dict((number, note) for number, note in enumerate(pitchnames))\n",
    "prediction = model.predict(network_input, verbose=0)\n",
    "print (numpy.argmax(prediction,axis=1))\n",
    "print ()\n",
    "predict_note= [int_to_note[index] for index in numpy.argmax(prediction,axis=1)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[70, 70, 46, 34, 36, 48, 67, 67, 67, 67, 67, 67, 67, 67, 69, 69, 70, 70, 69, 69, 36, 48, 50, 38, 67, 67, 67, 67, 70, 70, 69, 69, 70, 70, 69, 69, 67, 67, 50, 38, 43, 31, 67, 67, 67, 67, 69, 69, 70, 70, 43, 31, 46, 34, 67, 67, 67, 67, 69, 69, 70, 70, 46, 34, 41, 53, 67, 67, 67, 67, 67, 67, 67, 67, 69, 69, 70, 70, 69, 69, 41, 53, 67, 50, 38, 67, 67, 67, 50, 38, 53, 41, 70, 70, 69, 69, 70, 70, 69, 69, 67, 67, 53, 41]\n",
      "0.875\n"
     ]
    }
   ],
   "source": [
    "print (notes[sequence_length:])\n",
    "accurcy=np.sum([1 if x==y else 0 for x,y in zip(predict_note,notes[sequence_length:])])/len(predict_note)\n",
    "print (accurcy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "!open featureToMidi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import mido\n",
    "import numpy as np\n",
    "from mido import Message, MidiFile, MidiTrack\n",
    "import pickle\n",
    "import sys\n",
    "\n",
    "MAX_NUM_FILES = 20\n",
    "MIN_NUM_MESSAGES_PER_FILE = 100\n",
    "\n",
    "''' \n",
    "Returns array of song features.  Goes through first MAX_NUM_FILES in data directory\n",
    "Periodically saves the data to features.pickle\n",
    "'''\n",
    "def getAllFeatureVectors():\n",
    "    rootdir = os.getcwd() + '/data/'\n",
    "    \n",
    "    featureVectors = []\n",
    "    i = 0\n",
    "    for subdir, dirs, files in os.walk(rootdir):\n",
    "        for file in files:\n",
    "            #print i\n",
    "            if (i >= MAX_NUM_FILES):\n",
    "                break\n",
    "            \n",
    "            midiFileName = os.path.join(subdir, file)\n",
    "            #sys.stdout.write(midiFileName)\n",
    "            \n",
    "            #if (\"beatle\" not in midiFileName.lower()):\n",
    "            #    continue\n",
    "            if (not midiFileName.endswith('.mid')):\n",
    "                continue\n",
    "            #featureVector = getFeatureVector(midiFileName)\n",
    "            try:\n",
    "                featureVector = getFeatureVector(midiFileName)\n",
    "                #print 'len',len(featureVector)\n",
    "                if (len(featureVector) >= MIN_NUM_MESSAGES_PER_FILE):\n",
    "                    #print 'file number: ', os.path.getsize(midiFileName), i, file\n",
    "                    featureVectors.append(featureVector)\n",
    "                else:\n",
    "                    \n",
    "                    i += -1\n",
    "            except:\n",
    "                #print('exception')\n",
    "                i += -1\n",
    "            i+=1\n",
    "                \n",
    "            #save the data every so often \n",
    "            if (i %1000 == 0):\n",
    "                pickling_on = open(\"features.pickle\",\"wb\")\n",
    "                pickle.dump(np.array(featureVectors), pickling_on)\n",
    "                pickling_on.close()\n",
    "        if (i >= MAX_NUM_FILES):\n",
    "            break\n",
    "           \n",
    "   # Convert to numpy arrays\n",
    "    return np.array(featureVectors)\n",
    "\n",
    "\n",
    "def getFeatureVector(midiFileName):\n",
    "    file = MidiFile(midiFileName)\n",
    "    \n",
    "    #Get All the messages\n",
    "    messages = []\n",
    "    for message in file:#TODO: TRACKS???\n",
    "        if message.type in ['note_on','note_off']:\n",
    "            messages.append(message)\n",
    "    \n",
    "    features = []\n",
    "    note = None\n",
    "    duration = None\n",
    "    timeSincePreviousNote = 0.0\n",
    "    \n",
    "    i = 0\n",
    "    for message in messages:\n",
    "        try:\n",
    "            timeSincePreviousNote += message.time #TODO: DETERMINE IF THIS IS IN TICKS, IF SO , CHANGE TO SECONDS OR MAYBE NOT. IDK\n",
    "        except:\n",
    "            j = 4\n",
    "            \n",
    "        #determine if the message is the start of the note\n",
    "        if (message.type not in ['note_on']):\n",
    "            continue\n",
    "        #don't use the percussion channel\n",
    "        if (message.channel >9):\n",
    "            continue\n",
    "        note = message.note\n",
    "        \n",
    "        #Find out when the note is finished playing to calculate duration\n",
    "        '''duration = 0.0\n",
    "        for futureMessage in messages[i:]:\n",
    "            if (type(futureMessage.time) is float or type(futureMessage.time) is int):\n",
    "                duration += float(futureMessage.time)\n",
    "            if futureMessage.type in ['note_off']:\n",
    "                if futureMessage.note == note:\n",
    "                    break\n",
    "        '''\n",
    "        #newFeatureVector = [note, velocity, duration, timeSincePreviousNote, message.channel]\n",
    "        #noteVect = np.zeros(128)\n",
    "        #noteVect[note] = 1\n",
    "        \n",
    "        newFeatureVector = [note, timeSincePreviousNote, message.velocity]\n",
    "        features.append(newFeatureVector)\n",
    "        \n",
    "        timeSincePreviousNote = 0.0\n",
    "        i += 1\n",
    "       \n",
    "    return np.array(features)\n",
    "   \n",
    "featureVectors = getAllFeatureVectors()\n",
    "\n",
    "#print featureVectors[0]\n",
    "\n",
    "pickling_on = open(\"beatles_features.pickle\",\"wb\")\n",
    "pickle.dump(featureVectors, pickling_on)\n",
    "pickling_on.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 335,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 41.   0. 100.]\n",
      "[28.0, 29.0, 30.0, 32.0, 33.0, 34.0, 35.0, 36.0, 37.0, 38.0, 39.0, 40.0, 41.0, 42.0, 43.0, 44.0, 45.0, 46.0, 47.0, 48.0, 49.0, 50.0, 51.0, 52.0, 53.0, 54.0, 55.0, 56.0, 57.0, 58.0, 59.0, 60.0, 61.0, 62.0, 63.0, 64.0, 65.0, 66.0, 67.0, 68.0, 69.0, 70.0, 71.0, 72.0, 73.0, 74.0, 75.0, 76.0, 77.0, 78.0, 79.0, 80.0, 81.0, 82.0, 83.0, 84.0, 85.0, 86.0, 87.0, 88.0, 89.0, 90.0, 91.0]\n"
     ]
    }
   ],
   "source": [
    "features = np.concatenate(getAllFeatureVectors())\n",
    "print (features[0])\n",
    "notes=np.hsplit(features,3)[0].ravel()\n",
    "pitchnames = sorted(set(item for item in notes))\n",
    "print (pitchnames)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 392,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "63\n",
      "20\n",
      "{28.0: 0, 29.0: 1, 30.0: 2, 32.0: 3, 33.0: 4, 34.0: 5, 35.0: 6, 36.0: 7, 37.0: 8, 38.0: 9, 39.0: 10, 40.0: 11, 41.0: 12, 42.0: 13, 43.0: 14, 44.0: 15, 45.0: 16, 46.0: 17, 47.0: 18, 48.0: 19, 49.0: 20, 50.0: 21, 51.0: 22, 52.0: 23, 53.0: 24, 54.0: 25, 55.0: 26, 56.0: 27, 57.0: 28, 58.0: 29, 59.0: 30, 60.0: 31, 61.0: 32, 62.0: 33, 63.0: 34, 64.0: 35, 65.0: 36, 66.0: 37, 67.0: 38, 68.0: 39, 69.0: 40, 70.0: 41, 71.0: 42, 72.0: 43, 73.0: 44, 74.0: 45, 75.0: 46, 76.0: 47, 77.0: 48, 78.0: 49, 79.0: 50, 80.0: 51, 81.0: 52, 82.0: 53, 83.0: 54, 84.0: 55, 85.0: 56, 86.0: 57, 87.0: 58, 88.0: 59, 89.0: 60, 90.0: 61, 91.0: 62}\n",
      "5212\n",
      "(5212, 12, 1)\n",
      "[[0.326087 ]\n",
      " [0.1630435]\n",
      " [0.       ]\n",
      " [0.326087 ]\n",
      " [0.       ]\n",
      " [0.1630435]\n",
      " [0.       ]\n",
      " [0.       ]\n",
      " [0.652174 ]\n",
      " [0.326087 ]]\n"
     ]
    }
   ],
   "source": [
    "n_catagory=len(pitchnames)\n",
    "print (n_catagory)\n",
    "features = getAllFeatureVectors()\n",
    "print (len(features))\n",
    "sequence_length = 12\n",
    "# create a dictionary to map pitches to integers\n",
    "note_to_int = dict((note, number) for number, note in enumerate(pitchnames))\n",
    "print (note_to_int)\n",
    "notestotal=[]\n",
    "notes_input = []\n",
    "notes_output = []\n",
    "notes_comp=[]\n",
    "other_input = []\n",
    "for i in range(len(features)):\n",
    "    notes=np.hsplit(features[i],3)[0].ravel()\n",
    "    times=np.hsplit(features[i],3)[1].ravel()\n",
    "    #velocitys=np.hsplit(features[i],3)[2].ravel()\n",
    "    for j in range(0, len(notes) - sequence_length, 1):\n",
    "        sequence_in =  notes[j:j + sequence_length]\n",
    "        sequence_out = notes[j + sequence_length]\n",
    "        notes_input.append([note_to_int[char] for char in sequence_in])\n",
    "        notes_output.append(note_to_int[sequence_out])\n",
    "        notes_comp.append(note_to_int[sequence_out])\n",
    "        other_input.append(times[j])\n",
    "n_patterns = len(notes_input)\n",
    "print (n_patterns)\n",
    "# reshape the input into a format compatible with LSTM layers\n",
    "notes_input = numpy.reshape(notes_input, (n_patterns, sequence_length, 1))\n",
    "notes_output = keras.utils.to_categorical(notes_output,num_classes=n_catagory)\n",
    "notes_comp=np.array(notes_comp)\n",
    "other_input = numpy.reshape(other_input, (n_patterns, 1))\n",
    "print (notes_input.shape)\n",
    "print (other_input[100:110])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 397,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "5212/5212 [==============================] - 8s 1ms/step - loss: 3.4957 - main_output_loss: 3.4715 - aux_output_loss: 0.1207\n",
      "Epoch 2/100\n",
      "5212/5212 [==============================] - 3s 549us/step - loss: 3.2354 - main_output_loss: 3.2154 - aux_output_loss: 0.0999\n",
      "Epoch 3/100\n",
      "5212/5212 [==============================] - 3s 558us/step - loss: 3.0884 - main_output_loss: 3.0684 - aux_output_loss: 0.0996\n",
      "Epoch 4/100\n",
      "5212/5212 [==============================] - 3s 565us/step - loss: 2.9724 - main_output_loss: 2.9525 - aux_output_loss: 0.0993\n",
      "Epoch 5/100\n",
      "5212/5212 [==============================] - 3s 560us/step - loss: 2.8814 - main_output_loss: 2.8616 - aux_output_loss: 0.0991\n",
      "Epoch 6/100\n",
      "5212/5212 [==============================] - 3s 563us/step - loss: 2.8135 - main_output_loss: 2.7938 - aux_output_loss: 0.0988\n",
      "Epoch 7/100\n",
      "5212/5212 [==============================] - 3s 576us/step - loss: 2.7474 - main_output_loss: 2.7277 - aux_output_loss: 0.0987\n",
      "Epoch 8/100\n",
      "5212/5212 [==============================] - 3s 593us/step - loss: 2.6949 - main_output_loss: 2.6752 - aux_output_loss: 0.0985\n",
      "Epoch 9/100\n",
      "5212/5212 [==============================] - 3s 598us/step - loss: 2.6424 - main_output_loss: 2.6227 - aux_output_loss: 0.0984\n",
      "Epoch 10/100\n",
      "5212/5212 [==============================] - 3s 559us/step - loss: 2.5951 - main_output_loss: 2.5754 - aux_output_loss: 0.0983\n",
      "Epoch 11/100\n",
      "5212/5212 [==============================] - 3s 564us/step - loss: 2.5447 - main_output_loss: 2.5251 - aux_output_loss: 0.0980\n",
      "Epoch 12/100\n",
      "5212/5212 [==============================] - 3s 540us/step - loss: 2.4967 - main_output_loss: 2.4771 - aux_output_loss: 0.0980\n",
      "Epoch 13/100\n",
      "5212/5212 [==============================] - 3s 555us/step - loss: 2.4543 - main_output_loss: 2.4347 - aux_output_loss: 0.0981\n",
      "Epoch 14/100\n",
      "5212/5212 [==============================] - 3s 566us/step - loss: 2.4128 - main_output_loss: 2.3932 - aux_output_loss: 0.0980\n",
      "Epoch 15/100\n",
      "5212/5212 [==============================] - 3s 562us/step - loss: 2.3757 - main_output_loss: 2.3561 - aux_output_loss: 0.0978\n",
      "Epoch 16/100\n",
      "5212/5212 [==============================] - 3s 566us/step - loss: 2.3405 - main_output_loss: 2.3209 - aux_output_loss: 0.0977\n",
      "Epoch 17/100\n",
      "5212/5212 [==============================] - 3s 589us/step - loss: 2.3025 - main_output_loss: 2.2830 - aux_output_loss: 0.0975\n",
      "Epoch 18/100\n",
      "5212/5212 [==============================] - 3s 568us/step - loss: 2.2741 - main_output_loss: 2.2545 - aux_output_loss: 0.0976\n",
      "Epoch 19/100\n",
      "5212/5212 [==============================] - 3s 569us/step - loss: 2.2439 - main_output_loss: 2.2244 - aux_output_loss: 0.0976\n",
      "Epoch 20/100\n",
      "5212/5212 [==============================] - 3s 574us/step - loss: 2.2103 - main_output_loss: 2.1908 - aux_output_loss: 0.0972\n",
      "Epoch 21/100\n",
      "5212/5212 [==============================] - 3s 580us/step - loss: 2.1777 - main_output_loss: 2.1582 - aux_output_loss: 0.0975\n",
      "Epoch 22/100\n",
      "5212/5212 [==============================] - 3s 576us/step - loss: 2.1598 - main_output_loss: 2.1403 - aux_output_loss: 0.0975\n",
      "Epoch 23/100\n",
      "5212/5212 [==============================] - 3s 604us/step - loss: 2.1285 - main_output_loss: 2.1090 - aux_output_loss: 0.0972\n",
      "Epoch 24/100\n",
      "5212/5212 [==============================] - 3s 587us/step - loss: 2.1072 - main_output_loss: 2.0877 - aux_output_loss: 0.0972\n",
      "Epoch 25/100\n",
      "5212/5212 [==============================] - 3s 604us/step - loss: 2.0794 - main_output_loss: 2.0600 - aux_output_loss: 0.0971\n",
      "Epoch 26/100\n",
      "5212/5212 [==============================] - 3s 595us/step - loss: 2.0506 - main_output_loss: 2.0312 - aux_output_loss: 0.0970\n",
      "Epoch 27/100\n",
      "5212/5212 [==============================] - 3s 582us/step - loss: 2.0294 - main_output_loss: 2.0101 - aux_output_loss: 0.0967\n",
      "Epoch 28/100\n",
      "5212/5212 [==============================] - 3s 614us/step - loss: 2.0068 - main_output_loss: 1.9875 - aux_output_loss: 0.0967\n",
      "Epoch 29/100\n",
      "5212/5212 [==============================] - 3s 642us/step - loss: 1.9775 - main_output_loss: 1.9582 - aux_output_loss: 0.0965\n",
      "Epoch 30/100\n",
      "5212/5212 [==============================] - 3s 565us/step - loss: 1.9570 - main_output_loss: 1.9377 - aux_output_loss: 0.0966\n",
      "Epoch 31/100\n",
      "5212/5212 [==============================] - 3s 583us/step - loss: 1.9336 - main_output_loss: 1.9143 - aux_output_loss: 0.0962\n",
      "Epoch 32/100\n",
      "5212/5212 [==============================] - 3s 587us/step - loss: 1.9106 - main_output_loss: 1.8913 - aux_output_loss: 0.0963\n",
      "Epoch 33/100\n",
      "5212/5212 [==============================] - 3s 585us/step - loss: 1.8868 - main_output_loss: 1.8675 - aux_output_loss: 0.0964\n",
      "Epoch 34/100\n",
      "5212/5212 [==============================] - 3s 585us/step - loss: 1.8669 - main_output_loss: 1.8477 - aux_output_loss: 0.0959\n",
      "Epoch 35/100\n",
      "5212/5212 [==============================] - 3s 609us/step - loss: 1.8450 - main_output_loss: 1.8257 - aux_output_loss: 0.0963\n",
      "Epoch 36/100\n",
      "5212/5212 [==============================] - 3s 630us/step - loss: 1.8221 - main_output_loss: 1.8028 - aux_output_loss: 0.0963\n",
      "Epoch 37/100\n",
      "5212/5212 [==============================] - 3s 638us/step - loss: 1.8037 - main_output_loss: 1.7845 - aux_output_loss: 0.0959\n",
      "Epoch 38/100\n",
      "5212/5212 [==============================] - 3s 645us/step - loss: 1.7826 - main_output_loss: 1.7635 - aux_output_loss: 0.0958\n",
      "Epoch 39/100\n",
      "5212/5212 [==============================] - 3s 601us/step - loss: 1.7596 - main_output_loss: 1.7404 - aux_output_loss: 0.0958\n",
      "Epoch 40/100\n",
      "5212/5212 [==============================] - 3s 592us/step - loss: 1.7425 - main_output_loss: 1.7234 - aux_output_loss: 0.0956\n",
      "Epoch 41/100\n",
      "5212/5212 [==============================] - 3s 598us/step - loss: 1.7244 - main_output_loss: 1.7053 - aux_output_loss: 0.0957\n",
      "Epoch 42/100\n",
      "5212/5212 [==============================] - 3s 590us/step - loss: 1.7063 - main_output_loss: 1.6872 - aux_output_loss: 0.0958\n",
      "Epoch 43/100\n",
      "5212/5212 [==============================] - 3s 627us/step - loss: 1.6820 - main_output_loss: 1.6629 - aux_output_loss: 0.09540s - loss: 1.6731 - main_output_loss: 1.6549 - aux_output_loss: \n",
      "Epoch 44/100\n",
      "5212/5212 [==============================] - 3s 623us/step - loss: 1.6621 - main_output_loss: 1.6431 - aux_output_loss: 0.0952\n",
      "Epoch 45/100\n",
      "5212/5212 [==============================] - 3s 603us/step - loss: 1.6449 - main_output_loss: 1.6259 - aux_output_loss: 0.0951\n",
      "Epoch 46/100\n",
      "5212/5212 [==============================] - 3s 595us/step - loss: 1.6238 - main_output_loss: 1.6047 - aux_output_loss: 0.0956\n",
      "Epoch 47/100\n",
      "5212/5212 [==============================] - 3s 634us/step - loss: 1.6078 - main_output_loss: 1.5889 - aux_output_loss: 0.0948\n",
      "Epoch 48/100\n",
      "5212/5212 [==============================] - 3s 595us/step - loss: 1.5888 - main_output_loss: 1.5697 - aux_output_loss: 0.0952\n",
      "Epoch 49/100\n",
      "5212/5212 [==============================] - 3s 619us/step - loss: 1.5622 - main_output_loss: 1.5432 - aux_output_loss: 0.0954\n",
      "Epoch 50/100\n",
      "5212/5212 [==============================] - 3s 588us/step - loss: 1.5512 - main_output_loss: 1.5322 - aux_output_loss: 0.0953\n",
      "Epoch 51/100\n",
      "5212/5212 [==============================] - 3s 608us/step - loss: 1.5338 - main_output_loss: 1.5148 - aux_output_loss: 0.0952\n",
      "Epoch 52/100\n",
      "5212/5212 [==============================] - 3s 611us/step - loss: 1.5205 - main_output_loss: 1.5015 - aux_output_loss: 0.0952\n",
      "Epoch 53/100\n",
      "5212/5212 [==============================] - 3s 592us/step - loss: 1.4975 - main_output_loss: 1.4785 - aux_output_loss: 0.0949\n",
      "Epoch 54/100\n",
      "5212/5212 [==============================] - 3s 582us/step - loss: 1.4814 - main_output_loss: 1.4625 - aux_output_loss: 0.0948\n",
      "Epoch 55/100\n",
      "5212/5212 [==============================] - 3s 581us/step - loss: 1.4671 - main_output_loss: 1.4482 - aux_output_loss: 0.0946\n",
      "Epoch 56/100\n",
      "5212/5212 [==============================] - 3s 577us/step - loss: 1.4501 - main_output_loss: 1.4312 - aux_output_loss: 0.0946\n",
      "Epoch 57/100\n",
      "5212/5212 [==============================] - 3s 579us/step - loss: 1.4355 - main_output_loss: 1.4165 - aux_output_loss: 0.0951\n",
      "Epoch 58/100\n",
      "5212/5212 [==============================] - 3s 580us/step - loss: 1.4150 - main_output_loss: 1.3959 - aux_output_loss: 0.0952\n",
      "Epoch 59/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5212/5212 [==============================] - 3s 585us/step - loss: 1.4022 - main_output_loss: 1.3832 - aux_output_loss: 0.0948\n",
      "Epoch 60/100\n",
      "5212/5212 [==============================] - 3s 584us/step - loss: 1.3831 - main_output_loss: 1.3641 - aux_output_loss: 0.0950\n",
      "Epoch 61/100\n",
      "5212/5212 [==============================] - 3s 573us/step - loss: 1.3652 - main_output_loss: 1.3462 - aux_output_loss: 0.0951\n",
      "Epoch 62/100\n",
      "5212/5212 [==============================] - 3s 586us/step - loss: 1.3536 - main_output_loss: 1.3346 - aux_output_loss: 0.0950\n",
      "Epoch 63/100\n",
      "5212/5212 [==============================] - 3s 569us/step - loss: 1.3322 - main_output_loss: 1.3133 - aux_output_loss: 0.0949\n",
      "Epoch 64/100\n",
      "5212/5212 [==============================] - 3s 585us/step - loss: 1.3222 - main_output_loss: 1.3032 - aux_output_loss: 0.0950\n",
      "Epoch 65/100\n",
      "5212/5212 [==============================] - 3s 579us/step - loss: 1.3025 - main_output_loss: 1.2835 - aux_output_loss: 0.0949\n",
      "Epoch 66/100\n",
      "5212/5212 [==============================] - 3s 587us/step - loss: 1.2915 - main_output_loss: 1.2724 - aux_output_loss: 0.0952\n",
      "Epoch 67/100\n",
      "5212/5212 [==============================] - 3s 589us/step - loss: 1.2789 - main_output_loss: 1.2599 - aux_output_loss: 0.0949\n",
      "Epoch 68/100\n",
      "5212/5212 [==============================] - 3s 582us/step - loss: 1.2619 - main_output_loss: 1.2429 - aux_output_loss: 0.0948\n",
      "Epoch 69/100\n",
      "5212/5212 [==============================] - 3s 586us/step - loss: 1.2488 - main_output_loss: 1.2297 - aux_output_loss: 0.0951\n",
      "Epoch 70/100\n",
      "5212/5212 [==============================] - 3s 583us/step - loss: 1.2335 - main_output_loss: 1.2145 - aux_output_loss: 0.0948\n",
      "Epoch 71/100\n",
      "5212/5212 [==============================] - 3s 582us/step - loss: 1.2246 - main_output_loss: 1.2057 - aux_output_loss: 0.0948\n",
      "Epoch 72/100\n",
      "5212/5212 [==============================] - 3s 586us/step - loss: 1.2106 - main_output_loss: 1.1916 - aux_output_loss: 0.0952\n",
      "Epoch 73/100\n",
      "5212/5212 [==============================] - 3s 586us/step - loss: 1.1974 - main_output_loss: 1.1785 - aux_output_loss: 0.0945\n",
      "Epoch 74/100\n",
      "5212/5212 [==============================] - 3s 587us/step - loss: 1.1771 - main_output_loss: 1.1582 - aux_output_loss: 0.0943\n",
      "Epoch 75/100\n",
      "5212/5212 [==============================] - 3s 583us/step - loss: 1.1639 - main_output_loss: 1.1449 - aux_output_loss: 0.0949\n",
      "Epoch 76/100\n",
      "5212/5212 [==============================] - 3s 584us/step - loss: 1.1519 - main_output_loss: 1.1330 - aux_output_loss: 0.0944\n",
      "Epoch 77/100\n",
      "5212/5212 [==============================] - 3s 584us/step - loss: 1.1410 - main_output_loss: 1.1222 - aux_output_loss: 0.0942\n",
      "Epoch 78/100\n",
      "5212/5212 [==============================] - 3s 585us/step - loss: 1.1227 - main_output_loss: 1.1039 - aux_output_loss: 0.0940\n",
      "Epoch 79/100\n",
      "5212/5212 [==============================] - 3s 586us/step - loss: 1.1155 - main_output_loss: 1.0967 - aux_output_loss: 0.0944\n",
      "Epoch 80/100\n",
      "5212/5212 [==============================] - 3s 581us/step - loss: 1.0988 - main_output_loss: 1.0800 - aux_output_loss: 0.0941\n",
      "Epoch 81/100\n",
      "5212/5212 [==============================] - 3s 586us/step - loss: 1.0854 - main_output_loss: 1.0665 - aux_output_loss: 0.0942\n",
      "Epoch 82/100\n",
      "5212/5212 [==============================] - 3s 587us/step - loss: 1.0768 - main_output_loss: 1.0580 - aux_output_loss: 0.0941\n",
      "Epoch 83/100\n",
      "5212/5212 [==============================] - 3s 586us/step - loss: 1.0649 - main_output_loss: 1.0461 - aux_output_loss: 0.0940\n",
      "Epoch 84/100\n",
      "5212/5212 [==============================] - 3s 578us/step - loss: 1.0551 - main_output_loss: 1.0363 - aux_output_loss: 0.0938\n",
      "Epoch 85/100\n",
      "5212/5212 [==============================] - 3s 586us/step - loss: 1.0415 - main_output_loss: 1.0228 - aux_output_loss: 0.0938\n",
      "Epoch 86/100\n",
      "5212/5212 [==============================] - 3s 579us/step - loss: 1.0272 - main_output_loss: 1.0084 - aux_output_loss: 0.0936\n",
      "Epoch 87/100\n",
      "5212/5212 [==============================] - 3s 582us/step - loss: 1.0201 - main_output_loss: 1.0014 - aux_output_loss: 0.0936\n",
      "Epoch 88/100\n",
      "5212/5212 [==============================] - 3s 585us/step - loss: 1.0086 - main_output_loss: 0.9899 - aux_output_loss: 0.0934\n",
      "Epoch 89/100\n",
      "5212/5212 [==============================] - 3s 588us/step - loss: 0.9999 - main_output_loss: 0.9814 - aux_output_loss: 0.0928\n",
      "Epoch 90/100\n",
      "5212/5212 [==============================] - 3s 585us/step - loss: 0.9867 - main_output_loss: 0.9679 - aux_output_loss: 0.0939\n",
      "Epoch 91/100\n",
      "5212/5212 [==============================] - 3s 582us/step - loss: 0.9728 - main_output_loss: 0.9542 - aux_output_loss: 0.0929\n",
      "Epoch 92/100\n",
      "5212/5212 [==============================] - 3s 586us/step - loss: 0.9657 - main_output_loss: 0.9472 - aux_output_loss: 0.0926\n",
      "Epoch 93/100\n",
      "5212/5212 [==============================] - 3s 586us/step - loss: 0.9564 - main_output_loss: 0.9378 - aux_output_loss: 0.0931\n",
      "Epoch 94/100\n",
      "5212/5212 [==============================] - 3s 587us/step - loss: 0.9458 - main_output_loss: 0.9272 - aux_output_loss: 0.0931\n",
      "Epoch 95/100\n",
      "5212/5212 [==============================] - 3s 584us/step - loss: 0.9311 - main_output_loss: 0.9124 - aux_output_loss: 0.0931\n",
      "Epoch 96/100\n",
      "5212/5212 [==============================] - 3s 583us/step - loss: 0.9219 - main_output_loss: 0.9033 - aux_output_loss: 0.0930\n",
      "Epoch 97/100\n",
      "5212/5212 [==============================] - 3s 583us/step - loss: 0.9151 - main_output_loss: 0.8965 - aux_output_loss: 0.0927\n",
      "Epoch 98/100\n",
      "5212/5212 [==============================] - 3s 584us/step - loss: 0.9050 - main_output_loss: 0.8864 - aux_output_loss: 0.0930\n",
      "Epoch 99/100\n",
      "5212/5212 [==============================] - 3s 587us/step - loss: 0.8999 - main_output_loss: 0.8813 - aux_output_loss: 0.0928\n",
      "Epoch 100/100\n",
      "5212/5212 [==============================] - 3s 586us/step - loss: 0.8826 - main_output_loss: 0.8640 - aux_output_loss: 0.0926\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x1a4bc9ee80>"
      ]
     },
     "execution_count": 397,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "main_input = Input(shape=(sequence_length,1), name='main_input')\n",
    "#aux_input = Input(shape=(1,), name='aux_input')\n",
    "#x = Embedding(output_dim=512, input_dim=(n_catagory,2), input_length=100)(main_input)\n",
    "lstm_out1 = LSTM(64,return_sequences=True)(main_input)\n",
    "lstm_out2 = LSTM(32)(lstm_out1)\n",
    "main_output = Dense(n_catagory, activation='softmax', name='main_output')(lstm_out2)\n",
    "#x = keras.layers.concatenate([main_output, aux_input])\n",
    "x = Dense(32, activation='relu')(main_output)\n",
    "x = Dense(32, activation='relu')(x)\n",
    "aux_output = Dense(1, activation='sigmoid', name='aux_output')(x)\n",
    "model = Model(inputs=[main_input], outputs=[main_output,aux_output])\n",
    "model.compile(optimizer='rmsprop',\\\n",
    "              loss={'main_output': 'categorical_crossentropy', 'aux_output': 'mean_squared_error'},\\\n",
    "              loss_weights={'main_output': 1., 'aux_output': 0.2})\n",
    "\n",
    "# And trained it via:\n",
    "\n",
    "model.fit({'main_input': notes_input},\\\n",
    "          {'main_output': notes_output, 'aux_output': other_input},\\\n",
    "          epochs=100, batch_size=32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 398,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5212, 63)\n",
      "[[0.13775192]\n",
      " [0.1873982 ]\n",
      " [0.13846077]\n",
      " [0.19238305]\n",
      " [0.26770246]\n",
      " [0.09019112]\n",
      " [0.0312456 ]\n",
      " [0.06561729]\n",
      " [0.17091355]\n",
      " [0.15939069]]\n",
      "62\n",
      "0.7603607060629317\n"
     ]
    }
   ],
   "source": [
    "int_to_note = dict((number, note) for number, note in enumerate(pitchnames))\n",
    "prediction1,prediction2 = model.predict({'main_input': notes_input})\n",
    "print (prediction1.shape)\n",
    "print (prediction2[:10])\n",
    "#print (numpy.argmax(prediction,axis=1))\n",
    "#print ()\n",
    "print (max(notes_comp))\n",
    "predict_note= [int_to_note[int(index)] for index in numpy.argmax(prediction1,axis=1)]\n",
    "notes_true=[int_to_note[int(index)] for index in notes_comp]\n",
    "accurcy=np.sum([1 if x==y else 0 for x,y in zip(predict_note,notes_true)])/len(predict_note)\n",
    "print (accurcy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import StandardScaler\n",
    "scaler = StandardScaler()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(5452, 3)"
      ]
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "features = np.concatenate(getAllFeatureVectors())\n",
    "features = scaler.fit_transform(features)\n",
    "features.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 303,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((2000, 6, 3), (2000, 3))"
      ]
     },
     "execution_count": 303,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def create_dataset(dataset, look_back=3):\n",
    "    dataX, dataY = [], []\n",
    "    for i in range(len(dataset)-look_back-1):\n",
    "        a = dataset[i:(i+look_back)]\n",
    "        dataX.append(a)\n",
    "        dataY.append(dataset[i + look_back])\n",
    "    return numpy.array(dataX), numpy.array(dataY)\n",
    "look_back = 6\n",
    "X, Y = create_dataset(features,look_back)\n",
    "X_train=X[:2000]\n",
    "Y_train=Y[:2000]\n",
    "X_test=X[2000:4000]\n",
    "Y_test=Y[2000:4000]\n",
    "X_train.shape,Y_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 304,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model = keras.Sequential()\n",
    "model.add(LSTM(64, input_shape=(look_back, 3)))\n",
    "model.add(Dense(32,activation='sigmoid'))\n",
    "model.add(Dense(32,activation='sigmoid'))\n",
    "model.add(Dense(3, activation='linear'))\n",
    "model.compile(loss='mean_squared_error', optimizer='sgd')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 306,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "2000/2000 [==============================] - 0s 216us/step - loss: 0.9526\n",
      "Epoch 2/100\n",
      "2000/2000 [==============================] - 0s 192us/step - loss: 0.9472\n",
      "Epoch 3/100\n",
      "2000/2000 [==============================] - 0s 192us/step - loss: 0.9412\n",
      "Epoch 4/100\n",
      "2000/2000 [==============================] - 0s 194us/step - loss: 0.9343\n",
      "Epoch 5/100\n",
      "2000/2000 [==============================] - 0s 194us/step - loss: 0.9265\n",
      "Epoch 6/100\n",
      "2000/2000 [==============================] - 0s 195us/step - loss: 0.9178\n",
      "Epoch 7/100\n",
      "2000/2000 [==============================] - 0s 192us/step - loss: 0.9078\n",
      "Epoch 8/100\n",
      "2000/2000 [==============================] - 0s 194us/step - loss: 0.8965\n",
      "Epoch 9/100\n",
      "2000/2000 [==============================] - 0s 195us/step - loss: 0.8835\n",
      "Epoch 10/100\n",
      "2000/2000 [==============================] - 0s 195us/step - loss: 0.8694\n",
      "Epoch 11/100\n",
      "2000/2000 [==============================] - 0s 197us/step - loss: 0.8540\n",
      "Epoch 12/100\n",
      "2000/2000 [==============================] - 0s 193us/step - loss: 0.8377\n",
      "Epoch 13/100\n",
      "2000/2000 [==============================] - 0s 195us/step - loss: 0.8193\n",
      "Epoch 14/100\n",
      "2000/2000 [==============================] - 0s 201us/step - loss: 0.8001\n",
      "Epoch 15/100\n",
      "2000/2000 [==============================] - 0s 190us/step - loss: 0.7813\n",
      "Epoch 16/100\n",
      "2000/2000 [==============================] - 0s 195us/step - loss: 0.7632\n",
      "Epoch 17/100\n",
      "2000/2000 [==============================] - 0s 196us/step - loss: 0.7447\n",
      "Epoch 18/100\n",
      "2000/2000 [==============================] - 0s 196us/step - loss: 0.7289\n",
      "Epoch 19/100\n",
      "2000/2000 [==============================] - 0s 196us/step - loss: 0.7152\n",
      "Epoch 20/100\n",
      "2000/2000 [==============================] - 0s 199us/step - loss: 0.7029\n",
      "Epoch 21/100\n",
      "2000/2000 [==============================] - 0s 199us/step - loss: 0.6942\n",
      "Epoch 22/100\n",
      "2000/2000 [==============================] - 0s 195us/step - loss: 0.6873\n",
      "Epoch 23/100\n",
      "2000/2000 [==============================] - 0s 198us/step - loss: 0.6821\n",
      "Epoch 24/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6784\n",
      "Epoch 25/100\n",
      "2000/2000 [==============================] - 0s 198us/step - loss: 0.6756\n",
      "Epoch 26/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6738\n",
      "Epoch 27/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6729\n",
      "Epoch 28/100\n",
      "2000/2000 [==============================] - 0s 192us/step - loss: 0.6716\n",
      "Epoch 29/100\n",
      "2000/2000 [==============================] - 0s 199us/step - loss: 0.6707\n",
      "Epoch 30/100\n",
      "2000/2000 [==============================] - 0s 199us/step - loss: 0.6701\n",
      "Epoch 31/100\n",
      "2000/2000 [==============================] - 0s 199us/step - loss: 0.6695\n",
      "Epoch 32/100\n",
      "2000/2000 [==============================] - 0s 198us/step - loss: 0.6689\n",
      "Epoch 33/100\n",
      "2000/2000 [==============================] - 0s 198us/step - loss: 0.6692\n",
      "Epoch 34/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6684\n",
      "Epoch 35/100\n",
      "2000/2000 [==============================] - 0s 199us/step - loss: 0.6682\n",
      "Epoch 36/100\n",
      "2000/2000 [==============================] - 0s 199us/step - loss: 0.6684\n",
      "Epoch 37/100\n",
      "2000/2000 [==============================] - 0s 199us/step - loss: 0.6681\n",
      "Epoch 38/100\n",
      "2000/2000 [==============================] - 0s 193us/step - loss: 0.6674\n",
      "Epoch 39/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6673\n",
      "Epoch 40/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6675\n",
      "Epoch 41/100\n",
      "2000/2000 [==============================] - 0s 199us/step - loss: 0.6668\n",
      "Epoch 42/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6669\n",
      "Epoch 43/100\n",
      "2000/2000 [==============================] - 0s 202us/step - loss: 0.6663\n",
      "Epoch 44/100\n",
      "2000/2000 [==============================] - 0s 206us/step - loss: 0.6666\n",
      "Epoch 45/100\n",
      "2000/2000 [==============================] - 0s 202us/step - loss: 0.6666\n",
      "Epoch 46/100\n",
      "2000/2000 [==============================] - 0s 201us/step - loss: 0.6664\n",
      "Epoch 47/100\n",
      "2000/2000 [==============================] - 0s 202us/step - loss: 0.6661\n",
      "Epoch 48/100\n",
      "2000/2000 [==============================] - 0s 202us/step - loss: 0.6660\n",
      "Epoch 49/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6657\n",
      "Epoch 50/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6659\n",
      "Epoch 51/100\n",
      "2000/2000 [==============================] - 0s 201us/step - loss: 0.6657\n",
      "Epoch 52/100\n",
      "2000/2000 [==============================] - 0s 196us/step - loss: 0.6655\n",
      "Epoch 53/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6655\n",
      "Epoch 54/100\n",
      "2000/2000 [==============================] - 0s 201us/step - loss: 0.6652\n",
      "Epoch 55/100\n",
      "2000/2000 [==============================] - 0s 201us/step - loss: 0.6651\n",
      "Epoch 56/100\n",
      "2000/2000 [==============================] - 0s 201us/step - loss: 0.6647\n",
      "Epoch 57/100\n",
      "2000/2000 [==============================] - 0s 199us/step - loss: 0.6653\n",
      "Epoch 58/100\n",
      "2000/2000 [==============================] - 0s 201us/step - loss: 0.6650\n",
      "Epoch 59/100\n",
      "2000/2000 [==============================] - 0s 201us/step - loss: 0.6654\n",
      "Epoch 60/100\n",
      "2000/2000 [==============================] - 0s 201us/step - loss: 0.6649\n",
      "Epoch 61/100\n",
      "2000/2000 [==============================] - 0s 201us/step - loss: 0.6648\n",
      "Epoch 62/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6647\n",
      "Epoch 63/100\n",
      "2000/2000 [==============================] - 0s 202us/step - loss: 0.6647\n",
      "Epoch 64/100\n",
      "2000/2000 [==============================] - 0s 205us/step - loss: 0.6647\n",
      "Epoch 65/100\n",
      "2000/2000 [==============================] - 0s 206us/step - loss: 0.6648\n",
      "Epoch 66/100\n",
      "2000/2000 [==============================] - 0s 203us/step - loss: 0.6645\n",
      "Epoch 67/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6644\n",
      "Epoch 68/100\n",
      "2000/2000 [==============================] - 0s 202us/step - loss: 0.6643\n",
      "Epoch 69/100\n",
      "2000/2000 [==============================] - 0s 203us/step - loss: 0.6644\n",
      "Epoch 70/100\n",
      "2000/2000 [==============================] - 0s 193us/step - loss: 0.6639\n",
      "Epoch 71/100\n",
      "2000/2000 [==============================] - 0s 205us/step - loss: 0.6642\n",
      "Epoch 72/100\n",
      "2000/2000 [==============================] - 0s 199us/step - loss: 0.6644\n",
      "Epoch 73/100\n",
      "2000/2000 [==============================] - 0s 203us/step - loss: 0.6646\n",
      "Epoch 74/100\n",
      "2000/2000 [==============================] - 0s 198us/step - loss: 0.6640\n",
      "Epoch 75/100\n",
      "2000/2000 [==============================] - 0s 201us/step - loss: 0.6646\n",
      "Epoch 76/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6643\n",
      "Epoch 77/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6643\n",
      "Epoch 78/100\n",
      "2000/2000 [==============================] - 0s 195us/step - loss: 0.6637\n",
      "Epoch 79/100\n",
      "2000/2000 [==============================] - 0s 185us/step - loss: 0.6640\n",
      "Epoch 80/100\n",
      "2000/2000 [==============================] - 0s 191us/step - loss: 0.6641\n",
      "Epoch 81/100\n",
      "2000/2000 [==============================] - 0s 183us/step - loss: 0.6639\n",
      "Epoch 82/100\n",
      "2000/2000 [==============================] - 0s 198us/step - loss: 0.6636\n",
      "Epoch 83/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6634\n",
      "Epoch 84/100\n",
      "2000/2000 [==============================] - 0s 202us/step - loss: 0.6639\n",
      "Epoch 85/100\n",
      "2000/2000 [==============================] - 0s 197us/step - loss: 0.6639\n",
      "Epoch 86/100\n",
      "2000/2000 [==============================] - 0s 201us/step - loss: 0.6635\n",
      "Epoch 87/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6639\n",
      "Epoch 88/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6635\n",
      "Epoch 89/100\n",
      "2000/2000 [==============================] - 0s 201us/step - loss: 0.6638\n",
      "Epoch 90/100\n",
      "2000/2000 [==============================] - 0s 199us/step - loss: 0.6635\n",
      "Epoch 91/100\n",
      "2000/2000 [==============================] - 0s 202us/step - loss: 0.6634\n",
      "Epoch 92/100\n",
      "2000/2000 [==============================] - 0s 196us/step - loss: 0.6638\n",
      "Epoch 93/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6633\n",
      "Epoch 94/100\n",
      "2000/2000 [==============================] - 0s 202us/step - loss: 0.6632\n",
      "Epoch 95/100\n",
      "2000/2000 [==============================] - 0s 199us/step - loss: 0.6631\n",
      "Epoch 96/100\n",
      "2000/2000 [==============================] - 0s 201us/step - loss: 0.6633 0s - loss: 0\n",
      "Epoch 97/100\n",
      "2000/2000 [==============================] - 0s 199us/step - loss: 0.6637\n",
      "Epoch 98/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6630\n",
      "Epoch 99/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6630 0s - loss:\n",
      "Epoch 100/100\n",
      "2000/2000 [==============================] - 0s 200us/step - loss: 0.6628\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x1a3f26e080>"
      ]
     },
     "execution_count": 306,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X_train, Y_train, epochs=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 402,
   "metadata": {},
   "outputs": [
    {
     "ename": "ImportError",
     "evalue": "Failed to import pydot. You must install pydot and graphviz for `pydotprint` to work.",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mAttributeError\u001b[0m                            Traceback (most recent call last)",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/keras/utils/vis_utils.py\u001b[0m in \u001b[0;36m_check_pydot\u001b[0;34m()\u001b[0m\n\u001b[1;32m     26\u001b[0m         \u001b[0;31m# to check the pydot/graphviz installation.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 27\u001b[0;31m         \u001b[0mpydot\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mDot\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcreate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mpydot\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mDot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     28\u001b[0m     \u001b[0;32mexcept\u001b[0m \u001b[0mException\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mAttributeError\u001b[0m: 'NoneType' object has no attribute 'Dot'",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[0;31mImportError\u001b[0m                               Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-402-437440fcd987>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mkeras\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mutils\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mplot_model\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0mplot_model\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mto_file\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'model.png'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/keras/utils/vis_utils.py\u001b[0m in \u001b[0;36mplot_model\u001b[0;34m(model, to_file, show_shapes, show_layer_names, rankdir)\u001b[0m\n\u001b[1;32m    133\u001b[0m             \u001b[0;34m'LR'\u001b[0m \u001b[0mcreates\u001b[0m \u001b[0ma\u001b[0m \u001b[0mhorizontal\u001b[0m \u001b[0mplot\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    134\u001b[0m     \"\"\"\n\u001b[0;32m--> 135\u001b[0;31m     \u001b[0mdot\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel_to_dot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mshow_shapes\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mshow_layer_names\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrankdir\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    136\u001b[0m     \u001b[0m_\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mextension\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mos\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msplitext\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mto_file\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    137\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mextension\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/keras/utils/vis_utils.py\u001b[0m in \u001b[0;36mmodel_to_dot\u001b[0;34m(model, show_shapes, show_layer_names, rankdir)\u001b[0m\n\u001b[1;32m     54\u001b[0m     \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodels\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mSequential\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     55\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 56\u001b[0;31m     \u001b[0m_check_pydot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     57\u001b[0m     \u001b[0mdot\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mpydot\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mDot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     58\u001b[0m     \u001b[0mdot\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mset\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'rankdir'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrankdir\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/keras/utils/vis_utils.py\u001b[0m in \u001b[0;36m_check_pydot\u001b[0;34m()\u001b[0m\n\u001b[1;32m     29\u001b[0m         \u001b[0;31m# pydot raises a generic Exception here,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     30\u001b[0m         \u001b[0;31m# so no specific class can be caught.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 31\u001b[0;31m         raise ImportError('Failed to import pydot. You must install pydot'\n\u001b[0m\u001b[1;32m     32\u001b[0m                           ' and graphviz for `pydotprint` to work.')\n\u001b[1;32m     33\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mImportError\u001b[0m: Failed to import pydot. You must install pydot and graphviz for `pydotprint` to work."
     ]
    }
   ],
   "source": [
    "from keras.utils import plot_model\n",
    "plot_model(model, to_file='model.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 307,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "Y_predict = model.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 308,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2000, 3)"
      ]
     },
     "execution_count": 308,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(Y_predict-Y_test).shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 309,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.063026936295108\n"
     ]
    }
   ],
   "source": [
    "test_loss=np.sum([np.dot(y_delta,y_delta) for y_delta in (Y_predict-Y_test)])/(Y_predict-Y_test).shape[0]/3\n",
    "print (test_loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 400,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1a4598b240>]"
      ]
     },
     "execution_count": 400,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAGypJREFUeJzt3Xt8XOV95/HPT5rRxZJl2ZZsg++ADTgsGFuACYRwS2oI\nIWmzbSBJk7TZ9WubyzZt075omiVZ2rya7qZJNvtKYNmWJKQJlM1uNoTQQsplmxBuMjZXx2DAxsbY\nli1bN1vSjOa3fzxnpJEsWyN5dHv8fb9e8zrnPOdo5nnk0fd5znPOjM3dERGRuJRNdgVERKT0FO4i\nIhFSuIuIREjhLiISIYW7iEiEFO4iIhFSuIuIREjhLiISIYW7iEiEUpP1wg0NDb5s2bLJenkRkWlp\n48aN+929caTjJi3cly1bRnNz82S9vIjItGRmO4o5TtMyIiIRUriLiERI4S4iEiGFu4hIhBTuIiIR\nUriLiERI4S4iEiGFu4iU3t4X4Ynb4OD2ya7JSWvSPsQkIpHpOgAv/Ag2/wDeejaUPfgFWPsxuOxP\nYeaCya3fSUbhLiJj15eBVx6EzT+Elx+AXAYWnAvr/waWvwOe/nvY+F3Y9AO4aANc8lmYMWeya31S\nMHeflBduampyff3ANLLzKdj/CixZB3NOA7PJrlFxsj1hFOkOc5ZDTeP0qftU9tZzIdCf/19weH/4\nvZ77QTjvRlhwzuBjW1+HR78Cz/0jVM6Et38G1v1BWJ9qcjl4/VHYvRne9v7wXp9izGyjuzeNeJzC\nXY7ryEH4+c3wzJ0DZTNPgaVvh6WXwLJLoWHl1AnM7nbY9RTseBzeeBze3AjZ7oH9FbUwe1l4zFkO\ns5eH5ZzToG4RlOtk9pg6W+D5e0Ko730Byitg5XpY/WE44yooTx//5/e+BI98GX59H8yYC5f+MVzw\nCUhXT0z9j6djD2z6h/A+P5T/6haDM6+Fiz8Z3utT5D2ucJcT4w4v/QTu/1M4fADe/mn4N78TgnP7\nY7DjMeh4Kxw7oyGE/bJLwx/BvFVQNkHX6jv3wY5fwRtPwBu/gj3Pg+fAyuGUc2HJ28PZRqoyjCAP\nvj6wPLgd+noHnqssBfVLBgK/cDl7GVTMmJg2TSXZnjDdsvmHYfrF++DUNbD6Q3DOB8Y2xfLmRnjo\nL+G1R2DmqfDOP4PzPzJy51BquT549eEwbbT1n0Lblr0D1nwMFl8Az3wfmu+AI61hqmndJ+Gc3wrv\npUmkcJexa98NP/scbP0ZnHIeXP/fw7KQO7S+FoJ1x2Mh8NveCPuq6gtG9peEP4yy8hOvl3sI5R2P\nhyDf8Ti0vhr2paphUVN43SUXw6ILoLL2+M+Xy0HH7hD2ra8NDv7W7dDTNvj42gVhhD80+OtOLU37\njqVq1sQGijvs3gTP3hWmXY4cDG0/74Nw3odg3lmleZ3XfwEP3RIGDLOXwxV/ETqM8R4YtL0ZRumb\nvg9tO8PgZPWHQqg3nDH42MyRMJ30xK3Q8muonQ8X/Dto+n2oaRjfeh6Dwl1GL5eDjd+Bf/lSuFB2\nxefDaKXYqYpDbwyM6nc8FgIToLIOFl8Ugn7ppXDq6uJGabk+2PfS4DDv3BP2VdWHEF96cRidn3Ie\npCrG1OxhuYdQGzrazy/zZy0TpXpOuNukdn6YFpuZLGvnh/KZC0IAp6vG/hode0KQbb4LWrZAeSWc\n9Z4QfKddMT5TVu7hjOChv4S9z4ezviu/EKZDSjkN0peFbT8Po/RXHgxnd6ddEe7kOfM9I7933MMo\n/4lvw7Z/Cb+bc38n/H3MX1W6ehZB4S6j0/Iy/PQ/hnnq0y6H675+4heT2ncPHtnv3xrK0zNg8YUh\n6JddAgvXhpFptgfefGYgyHc+NTB6rluUBPm6EOaNZ03c1M9weg+HudnWfNCP09+ROxxuDZ1ax97w\nWp17wyOXPfr4qlkFoZ90ArULBjqAfCeQn2LKdMPW+8O0y6sPhdBbdEEI9Lf9JlTPHp92DZXLwUs/\nhoe/HM7GFq6Fq24O78UTceiNML2y6fvhd1c7P0wBnf+74cxrLFq2hpH8s3dD9kjoJNZ9Es64ekLe\nkyULdzO7A7gO2Ofu5wyz/yzgO8Aa4C/c/avFVFDhPkVke+Gxb8C//tcQuuv/OtzxMB4XjzpbBkb1\nO34VLspBGAU1rIT9L0NfTyhrOHNgVL704jAXLgNyuXAtpHNPGHF37BmyvndgPZc5+ucr60LQd+6F\n7jaoWxjudln9IWhYMfHtyevLhumgR78C7btg+WVw5c1hDrzo58iEOfRnvgfbHgplZ1wdRukr15du\nbv9wazjTfep/ho5j7gpY9x/C309FTWleYxilDPfLgE7gzmOE+zxgKfB+4KDCfRrZ+TTc+5lwCn7O\nB2D9V6B23sS9/uHWcKaw/bEw/TL/bWGqZcnFUDN34uoRs/z0UsdbQzqB5CygohbO/W1Y/s7xvW4w\nWpnuMIXyi69CVwusvCZM1wy9zbJQ6+vhbpdN/wBd+8LF2jW/G0bq4zk4yPaGmw+e+Fa4VlFVD2s/\nDhdugFkLS/5yJZ2WMbNlwH3DhXvBMV8COqMI974M9HZB5nA4/c50DVkeHmZ/McccDrd9rVwPZ78X\nTr9ycu7A6OmAh/8KnvwfYcR23ddg5W9MfD1ERtLTCU/eBo99E3rawyDkis/D3NPD/mxvuLXyme/B\na4+ClcGK3wjhesbVE3trq3u4a+uJb4c6YeFe+XWfgkVrS/YyCveRZHvgwKtw4JUwHbD/lfA4sC28\niUajvDKEdLomnI71r88IUx0VNclyRrh1b+s/QfehUHbG1XD29bDy3WG+dLy9/CD87I+hbVcYWVz1\nn6bmh0lECh05GAL+ydvC3+75Hwl/L5t/GD5ENWtJGKWv/vC4jJZH7eD2MF3zzJ0hTxZfFD64ddZ7\nT7jDmZLhbmYbgA0AS5YsWbtjR1H/z+vYuUPX/iEBniwP7QgXj/LqFoW5xoYVUDNv+GAeNrBrRn86\n25eB7b+ELT8NPXzn3vCBkNMuDyP6M68t/W1WnS3wzzeF7/5oPCvc3rj4wtK+hsh469gLv/xauP/c\nc3DmNbDm43D6FVNrWimvuz18186TyZeozVocBlVrPgrV9WN6yikZ7oVKOnLvy4Rf3P6XB4/C978c\nRsh5qapw0aPhjHABr2FlCPO5Z4zrBZDjyuVg19Ow5d4Q9od2hFPLpZeEoD/ruhMbibiHC1QPfD5M\nE73jc3DpH5X2tkGRidbZEi76T9K95qOW6wtn7E/cCjt+Ge6Vf8/fjump4g33tl1hbm3/y7B/W1ge\nfH3wbWG1CwZG4f0BviL0mpN5+9xI3MMnLLf8NDxatoTyhWtD0J99/cBcYzFaX4f7Pht+X4vXwfXf\nhMYzx6XqIlKk3ZvDqH32sjH9eCnvlrkLuBxoAPYCXwTSAO5+m5ktAJqBOiBHuLNmlbsfd+J6zOH+\n0k/gno+GaYw5pw8T4mdMzNz1RNj/ysCIfvemUDbvbUnQvzfcXTLcLYt9WXjy1nDPcFkK3vUlWPv7\nU7tjE5GixPshpu72cGtU/dKT60ueDu0M8/NbfhruEcfDR7bzI/qFa0N4v/VcuL3xrc3h9rH3/O3U\nuMAkIiURb7hLcsfN/SHoX/t/4UMqM08J363y6/vDN+5d+19g1funzDfZiUhpFBvuJ9HQNyK188J9\nvGs/DkcOhW/t23JvuANn9Y3w7r+auI+Ni8iUpHCf7qrrk2/r++Bk10REphBdYRMRiZDCXUQkQgp3\nEZEIKdxFRCKkcBcRiZDCXUQkQgp3EZEIKdxFRCKkcBcRiZDCXUQkQgp3EZEIKdxFRCKkcBcRiZDC\nXUQkQgp3EZEIKdxFRCKkcBcRiZDCXUQkQgp3EZEIKdxFRCKkcBcRiZDCXUQkQgp3EZEIKdxFRCKk\ncBcRiZDCXUQkQgp3EZEIKdxFRCKkcBcRiZDCXUQkQgp3EZEIjRjuZnaHme0zsxeOsd/M7Jtmts3M\nnjOzNaWvpoiIjEYxI/fvAuuPs/8aYEXy2ADceuLVEhGREzFiuLv7vwKtxznkfcCdHjwB1JvZKaWq\noIiIjF4p5twXAjsLtnclZUcxsw1m1mxmzS0tLSV4aRERGU4pwt2GKfPhDnT32929yd2bGhsbS/DS\nIiIynFKE+y5gccH2ImB3CZ5XRETGqBThfi/w0eSumXVAm7u/VYLnFRGRMUqNdICZ3QVcDjSY2S7g\ni0AawN1vA+4HrgW2AYeB3xuvyoqISHFGDHd3v3GE/Q58qmQ1EhGRE6ZPqIqIREjhLiISIYW7iEiE\nFO4iIhFSuIuIREjhLiISIYW7iEiEFO4iIhFSuIuIREjhLiISIYW7iEiEFO4iIhFSuIuIREjhLiIS\nIYW7iEiEFO4iIhFSuIuIREjhLiISIYW7iEiEFO4iIhFSuIuIREjhLiISIYW7iEiEFO4iIhFSuIuI\nREjhLiISIYW7iEiEFO4iIhFSuIuIREjhLiISIYW7iEiEFO4iIhFSuIuIRKiocDez9Wa21cy2mdlN\nw+xfamYPmdlzZvaomS0qfVVFRKRYI4a7mZUD3wKuAVYBN5rZqiGHfRW4093PBW4B/rrUFRURkeIV\nM3K/ENjm7q+5ey9wN/C+IcesAh5K1h8ZZr+IiEygYsJ9IbCzYHtXUlboWeADyfpvAjPNbO7QJzKz\nDWbWbGbNLS0tY6mviIgUoZhwt2HKfMj254B3mtkm4J3Am0D2qB9yv93dm9y9qbGxcdSVFRGR4qSK\nOGYXsLhgexGwu/AAd98N/BaAmdUCH3D3tlJVUkRERqeYkfvTwAozW25mFcANwL2FB5hZg5nln+vP\ngTtKW00RERmNEcPd3bPAp4EHgC3APe7+opndYmbXJ4ddDmw1s5eB+cCXx6m+IiJSBHMfOn0+MZqa\nmry5uXlSXltEZLoys43u3jTScfqEqohIhBTuIiIRUriLiERI4S4iEiGFu4hIhBTuIiIRUriLiERI\n4S4iEiGFu4hIhBTuIiIRUriLiERI4S4iEiGFu4hIhBTuIiIRUriLiERI4S4iEiGFu4hIhBTuIiIR\nUriLiERI4S4iEiGFu4hIhBTuIiIRUriLiERI4S4iEiGFu4hIhBTuIiIRUriLiERI4S4iEiGFu4hI\nhBTuIiIRUriLiERI4S4iEiGFu4hIhIoKdzNbb2ZbzWybmd00zP4lZvaImW0ys+fM7NrSV1VERIo1\nYribWTnwLeAaYBVwo5mtGnLYF4B73P184Abg26WuqIiIFK+YkfuFwDZ3f83de4G7gfcNOcaBumR9\nFrC7dFUUEZHRShVxzEJgZ8H2LuCiIcd8CXjQzD4D1ABXl6R2IiIyJsWM3G2YMh+yfSPwXXdfBFwL\nfN/MjnpuM9tgZs1m1tzS0jL62oqISFGKCfddwOKC7UUcPe3yCeAeAHd/HKgCGoY+kbvf7u5N7t7U\n2Ng4thqLiMiIign3p4EVZrbczCoIF0zvHXLMG8BVAGZ2NiHcNTQXEZkkI4a7u2eBTwMPAFsId8W8\naGa3mNn1yWF/Avx7M3sWuAv4uLsPnboREZEJUswFVdz9fuD+IWU3F6y/BFxS2qqJiMhY6ROqIiIR\nUriLiERI4S4iEiGFu4hIhBTuIiIRUriLiERI4S4iEiGFu4hIhBTuIiIRUriLiERI4S4iEiGFu4hI\nhBTuIiIRUriLiERI4S4iEiGFu4hIhBTuIiIRUriLiERI4S4iEiGFu4hIhBTuIiIRUriLiERI4S4i\nEiGFu4hIhBTuIiIRUriLiERI4S4iEiGFu4hIhBTuIiIRUriLiERI4S4iEiGFu4hIhBTuIiIRKirc\nzWy9mW01s21mdtMw+79uZpuTx8tmdqj0VRURkWKlRjrAzMqBbwHvAnYBT5vZve7+Uv4Yd/+jguM/\nA5w/DnUVEZEiFTNyvxDY5u6vuXsvcDfwvuMcfyNwVykqJyIiY1NMuC8EdhZs70rKjmJmS4HlwMMn\nXjURERmrYsLdhinzYxx7A/Ajd+8b9onMNphZs5k1t7S0FFtHEREZpWLCfRewuGB7EbD7GMfewHGm\nZNz9dndvcvemxsbG4mspIiKjUky4Pw2sMLPlZlZBCPB7hx5kZmcCs4HHS1tFEREZrRHD3d2zwKeB\nB4AtwD3u/qKZ3WJm1xcceiNwt7sfa8pGREQmyIi3QgK4+/3A/UPKbh6y/aXSVUtERE6EPqEqIhIh\nhbuISIQU7iIiEVK4i4hESOEuIhIhhbuISIQU7iIiESrqPvep5Nmdh/jer7YzsypFXXWamVUpZlaF\nZV3VwHZdsr8yVYbZcF+PMz31ZPtoP5KloztDe3eW9iMZ2rszdHRn6erJUledZn5dFQuSR111Kqr2\ni0wH7k5HT5a2wxkOHu7lULJsO5LhYFeG85fUc9nK8f0KlmkX7vs7e3jy9VY6ujN09GQZ6fOw6XIb\nJvyP7hD6l9VpKlJllBmYGeVmlJlhBuVlYb3MoKxw3SzZTtaPUV5eZmRzHuo+JJjDesGyO0P7kWRf\nd6Y/0HuyuVH9vqrSZcyvqxoI/FlVzJtZyYJZYXt+XRXz6iqpTJWfwL+KTDW92RxtRzIFjxAsbYcz\ntB3J9pc7Ht4fMyv73wvzZk7N90Qu5xw6kqG1q4f9nb30ZnOkyo10eRmpsmRZbqTKykiXG6nysEyX\nlQ06rrzMih7wuDuHe/s4dCTDwa4knJOwPtQf2uH3e7Cg7NCRDH25Y4fTH1x++riHu03WtwU0NTV5\nc3PzCT1HLud09Wbp6M72h2B/cHYno9skFDu6B0a7A9tZOnuyJWpRaVSmysKZR/XgDqeuoKyu4Kwl\nlIf1msoUbYcz7GnvZm97N3vakmV7D3vbutnbEcqG6yDm1FQkHUBl0gFUDeoA5tdVUlOZoieboyfT\nF5bZProzYdmTydGTzdF9jH3dBccM2pf8TLbPmVWdZm5tBXNqKplbU8Gcmgrm1lYwt6aSObUVzK2p\noCo9tQKnlNydbM7J9jmZXI5sn5Pty9HeHcK4fVBYh8ehw5lh9x3JDPvFrP1qKsqZVZ3GzNjX0U2m\n7+gcmD0jnQR+GBDMr0s6gJnh/TCvrorG2koqUmOb3XV32ruztHb1cqAzBHZ+/UBXb3h09nCgM6wf\nPNx73MAcjXTSCQztHPKdAtDfGfb2HXtANaOinNkzKphVnWZ2TZr66grqZ6Spn5HuL6+fUcHspKw+\nKUuXj31G3Mw2unvTSMdNu5F7obKy/Kg8Pebn6Ms5nf0dQwj+TJ/T507OHXcnl4O+/LqHnwn7Bq/n\nPP9z4Y0b9g1eLzOOCua6gjOJEw2vuqo0i+fMOOZ+d6ftSL4DCKG/pz089iXL599s50BXz4hnRaNR\nXmZUpcqoTJdTmSqjMlVGVf96OenyMna3dfPC7jZau3qHDRsIoTRn2A5goCx0EKFTqK4Y/vfp7klH\nNdDZdGf76M4MdDzdmVyy3Teo4ypc5tezfU6mL0c2lyz7nGwuRyZZ5vdnksDO5MJyUJCPMrjyAV1X\nHYJjWcMMZlWnBz3qhmznywrDJT8i3psMCva194T1jmS9o4dX9nawr6Nn2HCdW1PBvGQAMD8Z9c+r\nq2JuTQWd3Vn2d4WAbu3qZX/nwPqBrp5j/jvPrErRUBv+PZfOncGapbNpyP87J+WVqbKjfr/D/f4z\nQ3/PBb//wT+ff44Q5vlgrq9OgjoJ7PoZaeqr08yakZ5yZzeFpvXIXcZPpi9HS0dP6ASSDuBwb19/\nIFcNF9LH2ZcaxUilcETX2jUwcgujuqSsfz08jjW6qk6XM6emglS5DQrunmxuzJ2XGf1tq0qVU5k+\nxrRAMjKsyJeXl5EuG5guGDpyzO9PpwY/X21lasSAngh9Oae1qzd0APngb+9JOoFksNDezf7OHob2\nATMqyvvPwvId8NzagvWaSubUVNBQW8nsmqkdmpPtpBi5y/hJl5dxan01p9ZXT/hrm1l/iC1vqBnx\neHensyc7qBMY2gHk3KlKlVOVDp1Q/iyiKp2UJfvCMSGw88cXdl5V6TIqyuO6SF+s8jKjcWYljTMr\ngVnHPC7bl+v/3ddVp457BiXjR+Eu057ZwPTcsiI6AxlfqfKBi/gyeXSfu4hIhBTuIiIRUriLiERI\n4S4iEiGFu4hIhBTuIiIRUriLiERI4S4iEqFJ+/oBM2sBdhQUNQD7J6UyE+tkaSecPG09WdoJJ09b\np3I7l7r7iF8pOWnhPpSZNRfzfQnT3cnSTjh52nqytBNOnrbG0E5Ny4iIREjhLiISoakU7rdPdgUm\nyMnSTjh52nqytBNOnrZO+3ZOmTl3EREpnak0chcRkRKZlHA3szvMbJ+ZvVBQNsfMfm5mryTL2ZNR\nt1Iys8Vm9oiZbTGzF83sD5PyqNpqZlVm9pSZPZu08z8n5cvN7Mmknf9oZhWTXddSMLNyM9tkZvcl\n27G2c7uZPW9mm82sOSmL6r2bZ2b1ZvYjM/t18vd68XRv62SN3L8LrB9SdhPwkLuvAB5Ktqe7LPAn\n7n42sA74lJmtIr629gBXuvt5wGpgvZmtA/4G+HrSzoPAJyaxjqX0h8CWgu1Y2wlwhbuvLrgtMLb3\nbt5/A/7Z3c8CziP8+07vtnryHz9P9ANYBrxQsL0VOCVZPwXYOll1G8c2/wR4V8xtBWYAzwAXET4E\nkkrKLwYemOz6laB9iwh/6FcC9wEWYzuTtmwHGoaURffeBeqA10muQcbS1qk05z7f3d8CSJbzJrk+\nJWVmy4DzgSeJsK3JVMVmYB/wc+BV4JC7Z5NDdgELJ6t+JfQN4M+A/P/IPZc42wngwINmttHMNiRl\n0b13gdOAFuA7yXTb35lZDdO8rVMp3KNlZrXA/wY+6+7tk12f8eDufe6+mjCyvRA4e7jDJrZWpWVm\n1wH73H1jYfEwh07rdha4xN3XANcQphQvm+wKjZMUsAa41d3PB7qYblMww5hK4b7XzE4BSJb7Jrk+\nJWFmaUKw/8Dd/09SHGVbAdz9EPAo4RpDvZnl/xP2RcDuyapXiVwCXG9m24G7CVMz3yC+dgLg7ruT\n5T7gx4ROO8b37i5gl7s/mWz/iBD207qtUync7wU+lqx/jDA/Pa2ZmQF/D2xx968V7IqqrWbWaGb1\nyXo1cDXhgtQjwL9NDpv27XT3P3f3Re6+DLgBeNjdP0xk7QQwsxozm5lfB94NvEBk710Ad98D7DSz\nM5Oiq4CXmOZtnZQPMZnZXcDlhG9e2wt8Efi/wD3AEuAN4LfdvXXCK1dCZnYp8AvgeQbmaD9PmHeP\npq1mdi7wPaCcMGC4x91vMbPTCCPcOcAm4CPu3jN5NS0dM7sc+Jy7XxdjO5M2/TjZTAE/dPcvm9lc\nInrv5pnZauDvgArgNeD3SN7LTNO26hOqIiIRmkrTMiIiUiIKdxGRCCncRUQipHAXEYmQwl1EJEIK\ndxGRCCncRUQipHAXEYnQ/wcJthTIR+L+KwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a4598b630>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#epoch=100\n",
    "n_neuron_onelayer=[12,16,20,24,28,32,36,40,44,48,52,56,60,64]\n",
    "train_loss_onelayer=[0.6684,0.6651,0.6646,0.6680,0.6622,0.6640,0.6594,0.6613,0.6624,0.6661,0.6605,\\\n",
    "            0.6652,0.6641,0.6681]\n",
    "test_loss_onelayer=[1.0919,1.1005,1.1029,1.0851,1.1206,1.1070,1.1202,1.1126,1.1126,1.1071,1.1278,\\\n",
    "          1.0988,1.1208,1.0964]\n",
    "plt.plot(n_neuron_onelayer,train_loss_onelayer)\n",
    "plt.plot(n_neuron_onelayer,test_loss_onelayer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 401,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1a4f40b8d0>]"
      ]
     },
     "execution_count": 401,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAD8CAYAAACb4nSYAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a4f4b68d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#epoch=100\n",
    "n_neuron_twolayer=[16,24,32,40,48,56,64]\n",
    "train_loss_twolayer=[0.6767,0.6756,0.6789,0.6780,0.6797,0.6800,0.6797]\n",
    "test_loss_twolayer=[1.0612,1.0717,1.0670,1.0636,1.0607,1.0700,1.0547]\n",
    "plt.plot(n_neuron_twolayer,train_loss_twolayer)\n",
    "plt.plot(n_neuron_twolayer,test_loss_twolayer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import seaborn as sns\n",
    "sns.set_style('darkgrid')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 0.7166391 , -0.12023745, -0.03724909],\n",
       "       [ 0.7168887 ,  0.05233854,  0.08632386],\n",
       "       [ 0.72470576,  0.08964669,  0.14873257],\n",
       "       ...,\n",
       "       [-0.34180415, -0.3976169 , -0.18748423],\n",
       "       [-0.360519  ,  0.36732563, -0.5765101 ],\n",
       "       [-0.22899117,  0.81024516, -0.4173819 ]], dtype=float32)"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Y_hat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x11795b358>,\n",
       " <matplotlib.lines.Line2D at 0x11795b4a8>,\n",
       " <matplotlib.lines.Line2D at 0x11795b630>]"
      ]
     },
     "execution_count": 69,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11777f0f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(scaler.inverse_transform(Y_hat),'.')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[6.9030693e+01, 9.6764714e-02, 9.9690567e+01],\n",
       "       [6.9034355e+01, 1.2288289e-01, 9.9964554e+01],\n",
       "       [6.9149033e+01, 1.2852922e-01, 1.0010293e+02],\n",
       "       ...,\n",
       "       [5.3502922e+01, 5.4785270e-02, 9.9357468e+01],\n",
       "       [5.3228367e+01, 1.7055398e-01, 9.8494926e+01],\n",
       "       [5.5157932e+01, 2.3758677e-01, 9.8847740e+01]], dtype=float32)"
      ]
     },
     "execution_count": 70,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scaler.inverse_transform(Y_hat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "ename": "IndexError",
     "evalue": "index 2 is out of bounds for axis 0 with size 2",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mIndexError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-32-fe4b36095a82>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     56\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mf\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mfeatureVectors\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     57\u001b[0m     \u001b[0;31m#print(i)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 58\u001b[0;31m     \u001b[0mfeaturesToMidi\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mf\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'beatles'\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0mstr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0;34m'.mid'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     59\u001b[0m     \u001b[0mi\u001b[0m\u001b[0;34m+=\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     60\u001b[0m \u001b[0;31m#featuresToMidi(featureVectors[44], 'testing.mid')\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-32-fe4b36095a82>\u001b[0m in \u001b[0;36mfeaturesToMidi\u001b[0;34m(features, fileName)\u001b[0m\n\u001b[1;32m     19\u001b[0m         \u001b[0mtimeFromStart\u001b[0m \u001b[0;34m+=\u001b[0m \u001b[0mdeltaTime\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     20\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 21\u001b[0;31m         \u001b[0mvelocity\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfeature\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     22\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     23\u001b[0m         \u001b[0mwithTimeFromStart\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0marray\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'note_on'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnote\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdeltaTime\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvelocity\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtimeFromStart\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mIndexError\u001b[0m: index 2 is out of bounds for axis 0 with size 2"
     ]
    }
   ],
   "source": [
    "from mido import Message, MidiFile, MidiTrack\n",
    "\n",
    "#turn the feature vector back into a MIDI file\n",
    "def featuresToMidi(features, fileName):\n",
    "    directory = \"featureToMidi\"\n",
    "    mid = MidiFile(type=0)\n",
    "    track = MidiTrack()\n",
    "    mid.tracks.append(track)\n",
    "    \n",
    "    #track.append(Message('program_change', program=12, time=0))\n",
    "    \n",
    "    withTimeFromStart = []\n",
    "    timeFromStart = 0\n",
    "    \n",
    "    for feature in features:\n",
    "        note = feature[0]\n",
    "        \n",
    "        deltaTime = feature[1]\n",
    "        timeFromStart += deltaTime\n",
    "        \n",
    "        velocity = feature[2]\n",
    "        \n",
    "        withTimeFromStart.append(np.array(['note_on', note, deltaTime, velocity, timeFromStart]))\n",
    "        #withTimeFromStart.append(np.array(['note_off', note, velocity, duration, timeFromStart+duration]))\n",
    "        \n",
    "    withTimeFromStart = np.array(withTimeFromStart)\n",
    "    \n",
    "    #sortedIndicies = withTimeFromStart[:,4].sort(axis=0)\n",
    "    #withTimeFromStart = withTimeFromStart[sortedIndicies][0]\n",
    "       \n",
    "    currTime = 0.0\n",
    "    for m in withTimeFromStart:\n",
    "        prevTime = currTime\n",
    "    \n",
    "        currTime = float(m[4])\n",
    "        \n",
    "        n = int(round(float(m[1])))\n",
    "        \n",
    "        t = currTime - prevTime\n",
    "        #TIME IS STORED IN TICKS NOT SECONDS.  SO IM GONNA JUST GUESS HERE\n",
    "        tempo = mido.bpm2tempo(128)\n",
    "        t = mido.second2tick(t, 500 , tempo)#second, ticks_per_beat, tempo)\n",
    "        #print currTime, t, m[0]\n",
    "        track.append(Message(type=m[0], note=n, velocity=int(round(float(m[3]))), channel=4, time=abs(int(round(t)))))\n",
    "    \n",
    "    cwd = os.getcwd()\n",
    "    rootdir = (cwd + '/' + directory + '/')\n",
    "    mid.save(directory + \"/\" + fileName)\n",
    "    \n",
    "    '''port = mido.open_output()\n",
    "    #port.send(msg)\n",
    "    for msg in MidiFile(directory + \"/\" + fileName).play():\n",
    "        port.send(msg)'''\n",
    "\n",
    "i = 0\n",
    "for f in featureVectors:\n",
    "    #print(i)\n",
    "    featuresToMidi(f, 'beatles'+str(i)+'.mid')\n",
    "    i+=1\n",
    "#featuresToMidi(featureVectors[44], 'testing.mid')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Functional form of loss function:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x = np.linspace(-48,47,96)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([-48., -47., -46., -45., -44., -43., -42., -41., -40., -39., -38.,\n",
       "       -37., -36., -35., -34., -33., -32., -31., -30., -29., -28., -27.,\n",
       "       -26., -25., -24., -23., -22., -21., -20., -19., -18., -17., -16.,\n",
       "       -15., -14., -13., -12., -11., -10.,  -9.,  -8.,  -7.,  -6.,  -5.,\n",
       "        -4.,  -3.,  -2.,  -1.,   0.,   1.,   2.,   3.,   4.,   5.,   6.,\n",
       "         7.,   8.,   9.,  10.,  11.,  12.,  13.,  14.,  15.,  16.,  17.,\n",
       "        18.,  19.,  20.,  21.,  22.,  23.,  24.,  25.,  26.,  27.,  28.,\n",
       "        29.,  30.,  31.,  32.,  33.,  34.,  35.,  36.,  37.,  38.,  39.,\n",
       "        40.,  41.,  42.,  43.,  44.,  45.,  46.,  47.])"
      ]
     },
     "execution_count": 88,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def loss_func(y_true, y_pred):\n",
    "    import scipy\n",
    "    x = np.linspace(-48,47,96)\n",
    "    y = np.array([0.8, 1.8, 1.8, 1.8, 1. , 1.8, 1.8, 1. , 1.8, 1.8, 1.8, 1.8, 0.6,\n",
    "       1.6, 1.6, 1.6, 0.8, 1.6, 1.6, 0.8, 1.6, 1.6, 1.6, 1.6, 0.4, 1.4,\n",
    "       1.4, 1.4, 0.6, 1.4, 1.4, 0.6, 1.4, 1.4, 1.4, 1.4, 0.2, 1.2, 1.2,\n",
    "       1.2, 0.4, 1.2, 1.2, 0.4, 1.2, 1.2, 1.2, 1.2, 0. , 1. , 1. , 1. ,\n",
    "       0.2, 1. , 1. , 0.2, 1. , 1. , 1. , 1. , 0.2, 1.2, 1.2, 1.2, 0.4,\n",
    "       1.2, 1.2, 0.4, 1.2, 1.2, 1.2, 1.2, 0.4, 1.4, 1.4, 1.4, 0.6, 1.4,\n",
    "       1.4, 0.6, 1.4, 1.4, 1.4, 1.4, 0.6, 1.6, 1.6, 1.6, 0.8, 1.6, 1.6,\n",
    "       0.8, 1.6, 1.6, 1.6, 1.6])\n",
    "    x = x[24:72]\n",
    "    y = y[24:72]\n",
    "    # fit = scipy.poly1d(scipy.polyfit(x,y,deg=40))\n",
    "    delta1 = y_true[0][0]-y_pred[0][0]\n",
    "    loss1 = scipy.poly1d(scipy.polyfit(x,y,deg=40))(delta1)\n",
    "    loss2 = (y_true[1]-y_pred[1])**2\n",
    "    loss3 = (y_true[2]-y_pred[2])**2\n",
    "    return loss1+loss2+loss3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y = [[0,1,1,1,0.2,1,1,0.2,1,1,1,1]]*8"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y = np.array(y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [
    {
     "ename": "IndexError",
     "evalue": "index 8 is out of bounds for axis 0 with size 8",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mIndexError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-120-3c5016fed748>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0my\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m8\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m8\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0;36m0.8\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mIndexError\u001b[0m: index 8 is out of bounds for axis 0 with size 8"
     ]
    }
   ],
   "source": [
    "y[8] = y[8]+0.8"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 172,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y = np.array([0.8, 1.8, 1.8, 1.8, 1. , 1.8, 1.8, 1. , 1.8, 1.8, 1.8, 1.8, 0.6,\n",
    "       1.6, 1.6, 1.6, 0.8, 1.6, 1.6, 0.8, 1.6, 1.6, 1.6, 1.6, 0.4, 1.4,\n",
    "       1.4, 1.4, 0.6, 1.4, 1.4, 0.6, 1.4, 1.4, 1.4, 1.4, 0.2, 1.2, 1.2,\n",
    "       1.2, 0.4, 1.2, 1.2, 0.4, 1.2, 1.2, 1.2, 1.2, 0. , 1. , 1. , 1. ,\n",
    "       0.2, 1. , 1. , 0.2, 1. , 1. , 1. , 1. , 0.2, 1.2, 1.2, 1.2, 0.4,\n",
    "       1.2, 1.2, 0.4, 1.2, 1.2, 1.2, 1.2, 0.4, 1.4, 1.4, 1.4, 0.6, 1.4,\n",
    "       1.4, 0.6, 1.4, 1.4, 1.4, 1.4, 0.6, 1.6, 1.6, 1.6, 0.8, 1.6, 1.6,\n",
    "       0.8, 1.6, 1.6, 1.6, 1.6])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import scipy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 189,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/Cellar/python3/3.6.3/Frameworks/Python.framework/Versions/3.6/lib/python3.6/site-packages/ipykernel_launcher.py:1: RankWarning: Polyfit may be poorly conditioned\n",
      "  \"\"\"Entry point for launching an IPython kernel.\n"
     ]
    }
   ],
   "source": [
    "fit = scipy.poly1d(scipy.polyfit(x,y,deg=40))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 175,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x = x[24:72]\n",
    "y = y[24:72]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 187,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "xp = np.linspace(-10,10,1000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 190,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1193c6940>]"
      ]
     },
     "execution_count": 190,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1179768d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(xp,fit(xp))#,x,y,'.')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/Cellar/python3/3.6.3/Frameworks/Python.framework/Versions/3.6/lib/python3.6/site-packages/ipykernel_launcher.py:1: RankWarning: Polyfit may be poorly conditioned\n",
      "  \"\"\"Entry point for launching an IPython kernel.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([-2.37821556e-57, -4.88059435e-56,  4.18184535e-53,  8.17973927e-52,\n",
       "       -3.23507804e-49, -5.96747024e-48,  1.41516494e-45,  2.40677790e-44,\n",
       "       -3.62046723e-42, -5.29968072e-41,  4.15538261e-39,  2.83200124e-38,\n",
       "        5.75891923e-36,  2.10335370e-34, -3.59151421e-32, -7.99291860e-31,\n",
       "        8.22885198e-29,  1.59910773e-27, -1.20350301e-25, -2.15605656e-24,\n",
       "        1.24214248e-22,  2.08957813e-21, -9.32556207e-20, -1.48607541e-18,\n",
       "        5.11605322e-17,  7.76689664e-16, -2.02847724e-14, -2.95038178e-13,\n",
       "        5.66851488e-12,  7.96737481e-11, -1.07012476e-09, -1.47823780e-08,\n",
       "        1.27565612e-07,  1.79480566e-06, -8.63225570e-06, -1.33005009e-04,\n",
       "        2.84517737e-04,  5.37000240e-03, -3.40385544e-03, -9.69933140e-02,\n",
       "        8.80846020e-01])"
      ]
     },
     "execution_count": 162,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scipy.polyfit(x,y,deg=40)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "        0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
       "        0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.]])"
      ]
     },
     "execution_count": 199,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scaler2.fit_transform([x])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.4, 1.4, 1.4, 1.4, 0.6, 1.4, 1.4, 0.6, 1.4, 1.4, 1.4, 1.4, 0.2,\n",
       "       1.2, 1.2, 1.2, 0.4, 1.2, 1.2, 0.4, 1.2, 1.2, 1.2, 1.2, 0. , 1. ,\n",
       "       1. , 1. , 0.2, 1. , 1. , 0.2, 1. , 1. , 1. , 1. , 0.2, 1.2, 1.2,\n",
       "       1.2, 0.4, 1.2, 1.2, 0.4, 1.2, 1.2, 1.2, 1.2])"
      ]
     },
     "execution_count": 200,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "scaler2 = StandardScaler()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 218,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.1421802099804372"
      ]
     },
     "execution_count": 218,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "fit(-6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}