06_tsp_cross_entropy.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Travelling Salesman Problem Reinforcement Learning - Cross Entropy\n",
    "\n",
    "The travelling salesman problem is a classic optimisation problem, where the aim is to find the fastest or shortest route between a numebr of cities, returning to the base city at the end."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### RL involves:\n",
    "* Trial and error search\n",
    "* Receiving and maximising reward (often delayed)\n",
    "* Linking state -> action -> reward\n",
    "* Must be able to sense something of their environment\n",
    "* Involves uncertainty in sensing and linking action to reward\n",
    "* Learning -> improved choice of actions over time\n",
    "* All models find a way to balance best predicted action vs. exploration\n",
    "\n",
    "### Elements of RL\n",
    "* *Environment*: all observable and unobservable information relevant to us\n",
    "* *Observation*: sensing the environment\n",
    "* *State*: the perceived (or perceivable) environment \n",
    "* *Agent*: senses environment, decides on action, receives and monitors rewards\n",
    "* *Action*: may be discrete (e.g. turn left) or continuous (accelerator pedal)\n",
    "* *Policy* (how to link state to action; often based on probabilities)\n",
    "* *Reward signal*: aim is to accumulate maximum reward over time\n",
    "* *Value function* of a state: prediction of likely/possible long-term reward from a particular state\n",
    "* *Q*: prediction of likely/possible long-term reward from performing a particular *action*\n",
    "* *Model* (optional): a simulation of the environment"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Cross Entropy\n",
    "\n",
    "Cross entropy is a policy method (it seeks to recommend actions without any estimation of the value of a state or action). The basic method is:\n",
    "\n",
    "1. Play *k* episodes using current model and environment\n",
    "2. Calculate total reward for each episode\n",
    "3. Keep best episodes (e.g. top 50%)\n",
    "4. Train on remaining episodes using observations and actions taken\n",
    "5. Repeat\n",
    "\n",
    "![](./images/cross_entropy.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import\n",
    "\n",
    "Import required liraries (including environment and plotting libraries in `env` folder)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from collections import deque\n",
    "from env.tsp_env import TspEnv\n",
    "from env import tsp_plots\n",
    "import numpy as np\n",
    "import time\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.optim as optim"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Set number of cities to visit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "NUMBER_OF_CITIES = 10"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Set stopping conditions\n",
    "\n",
    "Stop if any of the following are true:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "MAXIMUM_BATCHES = 5000\n",
    "MAXIMUM_TIME_MINS = 120\n",
    "NO_IMPROVEMENT_BATCHES = 250\n",
    "NO_IMPROVEMENT_TIME = 120"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Define neural net"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Net(nn.Module):\n",
    "    \"\"\"\n",
    "    Pytorch neural net using the flexible pytorch nn.Modlule class.\n",
    "    Note: the neural net output is linear. To convert these to probabilities for\n",
    "    each action (sum to 1.0) a SoftMax activation on the final output is\n",
    "    required, but this is applied outside of the net itself, which improves\n",
    "    speed and stability of training.\n",
    "\n",
    "    Layers in model:\n",
    "    * Input layer (implied, takes the number of observations)\n",
    "    * Densely connected layer (size = 4 x number of observations)\n",
    "    * ReLU activation\n",
    "    * Densely connected layer (size = 4 x number of observations)\n",
    "    * ReLU activation\n",
    "    * Output layer (size = number of possible actions)\n",
    "\n",
    "    (A SoftMax layer will be added later)\n",
    "    \"\"\"\n",
    "\n",
    "    def __init__(self, observation_space, action_space):\n",
    "        \"\"\"Define layers of sequential net\"\"\"\n",
    "        super(Net, self).__init__()\n",
    "        self.net = nn.Sequential(\n",
    "            nn.Linear(observation_space, NUMBER_OF_CITIES * 4),\n",
    "            nn.ReLU(),\n",
    "            nn.Linear(NUMBER_OF_CITIES * 4, NUMBER_OF_CITIES * 4),\n",
    "            nn.ReLU(),\n",
    "            nn.Linear(NUMBER_OF_CITIES * 4, NUMBER_OF_CITIES * 4),\n",
    "            nn.ReLU(),\n",
    "            nn.Linear(NUMBER_OF_CITIES * 4, action_space)\n",
    "        )\n",
    "        \n",
    "        \n",
    "    def forward(self, x):\n",
    "        \"\"\"Define forward pass (simple, as using a pre-defined sequential\n",
    "        model)\"\"\"\n",
    "        \n",
    "        # Move input to required device (GPU or CPU)\n",
    "        # Pass through net\n",
    "        return self.net(x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def play_episode(env, model, observation_space):\n",
    "    \"\"\"Play an episode\"\"\"\n",
    "\n",
    "    # Define softmax layer\n",
    "    sm = nn.Softmax(dim=1)\n",
    "\n",
    "    # Reset trackers and environment\n",
    "    episode_reward = 0\n",
    "    obs_tracker = []\n",
    "    action_tracker = []\n",
    "\n",
    "    # Reset environment (returns first observation)\n",
    "    obs, reward, is_terminal, info = env.reset()\n",
    "\n",
    "    # Loop up to 200 steps\n",
    "    for step in range(500):\n",
    "\n",
    "        # Track observations\n",
    "        obs = np.float32(obs)\n",
    "        obs_tracker.append(obs)\n",
    "\n",
    "        # Get action probability (put obs in Tensor first)\n",
    "        obs = torch.FloatTensor([obs])\n",
    "        act_probs = model(obs)\n",
    "        act_probs = sm(act_probs)\n",
    "        act_probs = act_probs.data.numpy()[0]\n",
    "\n",
    "        # Get and track action: action sampled based on probability distribution\n",
    "        np.random.seed()\n",
    "        action = np.random.choice(len(act_probs), p=act_probs)\n",
    "        action_tracker.append(action)\n",
    "\n",
    "        # Take action\n",
    "        obs, reward, done, info = env.step(action)\n",
    "        episode_reward += reward\n",
    "\n",
    "        # Pole has fallen over if done is True\n",
    "        if done:\n",
    "            break\n",
    "\n",
    "    # Put results in dictionary\n",
    "    results = {'episode_reward': episode_reward,\n",
    "               'episode_obs': obs_tracker,\n",
    "               'episode_actions': action_tracker}\n",
    "\n",
    "    return results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def filter_episodes(results, percentile_cutoff=70):\n",
    "    \"\"\"Get best episodes\"\"\"\n",
    "\n",
    "    # Get episode rewards & define cuttoff\n",
    "    episode_rewards = [episode['episode_reward'] for episode in results]\n",
    "    reward_cutoff = np.percentile(episode_rewards, percentile_cutoff)\n",
    "    \n",
    "    # Get best route\n",
    "    best_route_index = np.argmax(episode_rewards)\n",
    "    \n",
    "    all_actions = actions = [episode['episode_actions'] for episode in results]\n",
    "    best_route = all_actions[best_route_index]\n",
    "\n",
    "    # Get best episode observations and actions\n",
    "    obs = [episode['episode_obs'] for episode in results if\n",
    "           episode['episode_reward'] >= reward_cutoff]\n",
    "\n",
    "    actions = [episode['episode_actions'] for episode in results if\n",
    "               episode['episode_reward'] >= reward_cutoff]\n",
    "\n",
    "    # Convert list of observation arrays into a numpy array\n",
    "    obs = np.vstack(obs)\n",
    "\n",
    "    # Flatten actions list and convert to NumPy\n",
    "    actions = [item for sublist in actions for item in sublist]\n",
    "    actions = np.array(actions)\n",
    "    \n",
    "    return obs, actions, best_route"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def main():\n",
    "    \n",
    "    ############################## Set up model ################################\n",
    "\n",
    "    # Set number of episodes to play before slecting best and updating net\n",
    "    episode_batch_size = 50\n",
    "\n",
    "    # Set up environment\n",
    "    time_start = time.time()\n",
    "    env = TspEnv(number_of_cities = NUMBER_OF_CITIES)\n",
    "\n",
    "    # Get number of observations from environemt(allows the env to change)\n",
    "    # Obs = array of visited cities and on-ehot array of current city\n",
    "    obs_size = env.observation_space.shape[0] * 2\n",
    "\n",
    "    # Get number of actins from environemnt\n",
    "    n_actions = len(env.action_space)\n",
    "\n",
    "    # Set up Neural Net\n",
    "    model = Net(obs_size, n_actions)\n",
    "    objective = nn.CrossEntropyLoss()\n",
    "    learning_rate = 0.003\n",
    "    optimizer = optim.Adam(params=model.parameters(), lr=learning_rate)\n",
    "\n",
    "    # Play batches of episodes\n",
    "    batch_count = 0\n",
    "\n",
    "    # Lists to store results\n",
    "    all_results_batch_number = []\n",
    "    all_results_average_reward = []\n",
    "    all_results_maxumum_reward = []\n",
    "    last_results = deque(maxlen = 10)\n",
    "    \n",
    "    # Track best\n",
    "    best_reward = -9999999\n",
    "    best_average_reward = -9999999\n",
    "    best_route = None\n",
    "    \n",
    "    # Set batch and time of last best route\n",
    "    batch_last_best = 0\n",
    "    time_last_best = time.time()\n",
    "\n",
    "    ################## Play batch of epsiodes and select best ##################\n",
    "    \n",
    "    # Start playing loop and continue until goal performance reached\n",
    "    while True:\n",
    "        # Play episodes\n",
    "        batch_count += 1\n",
    "        batch_results = []\n",
    "        for episode in range(episode_batch_size):\n",
    "            results = play_episode(env, model, obs_size)\n",
    "            batch_results.append(results)\n",
    "\n",
    "        # Get average and maximum reward\n",
    "        rewards = [episode['episode_reward'] for episode in batch_results]\n",
    "        average_reward = np.mean(rewards)\n",
    "        maximum_reward = np.max(rewards)\n",
    "        \n",
    "        # Get best runs\n",
    "        training_obs, training_actions, batch_best_route = \\\n",
    "            filter_episodes(batch_results)\n",
    "        training_obs = torch.Tensor(training_obs)\n",
    "        training_actions = torch.Tensor(training_actions).long()        \n",
    "   \n",
    "        # Check whether best route is better than previously discovered\n",
    "        if maximum_reward > best_reward:\n",
    "            best_reward = maximum_reward\n",
    "            best_route = batch_best_route\n",
    "        \n",
    "        if average_reward > best_average_reward:\n",
    "            best_average_reward = average_reward\n",
    "            batch_last_best = batch_count\n",
    "            time_last_best = time.time()\n",
    "            time_elapsed = (time.time() - time_start) / 60\n",
    "            print(f'\\nNew best average. Batch : {batch_count: 5.0f},  ' \\\n",
    "                  f'Time {time_elapsed: 4.0f}  ' \\\n",
    "                  f'Average reward {average_reward: 6.0f}')\n",
    "\n",
    "        # Store and print results\n",
    "        all_results_batch_number.append(batch_count)\n",
    "        all_results_average_reward.append(average_reward)\n",
    "        all_results_maxumum_reward.append(maximum_reward)\n",
    "        last_results.append(average_reward)\n",
    "\n",
    "        print(f'\\rBatch {batch_count:3}. Average and best run: ' \\\n",
    "              f'{average_reward:6.0f}, {maximum_reward:6.0f}. ', end=\"\")\n",
    "        \n",
    "        # Check for stopping\n",
    "        stop = False\n",
    "        if batch_count == MAXIMUM_BATCHES:\n",
    "            stop = True\n",
    "        elif time.time() - time_start > MAXIMUM_TIME_MINS * 60:\n",
    "            stop = True\n",
    "        elif time.time() - time_last_best > NO_IMPROVEMENT_TIME * 60:\n",
    "            stop = True\n",
    "        elif batch_count - batch_last_best > NO_IMPROVEMENT_BATCHES:\n",
    "            stop = True\n",
    "        elif average_reward == maximum_reward:\n",
    "            # Model converged so all runs identical\n",
    "            stop = True\n",
    "  \n",
    "        if stop:\n",
    "            break \n",
    "\n",
    "        \n",
    "        ############### Train model on best episodes from batch ################\n",
    "       \n",
    "        # Reset model gradients\n",
    "        optimizer.zero_grad()\n",
    "        # Predict actions\n",
    "        action_scores = model(training_obs)\n",
    "        # Calculate loss between predicted and actual actions\n",
    "        loss_v = objective(action_scores, training_actions)\n",
    "        # Back propgate losses\n",
    "        loss_v.backward()\n",
    "        # Update model weights\n",
    "        optimizer.step()\n",
    "        \n",
    "    ############################# Plot results #################################\n",
    "        \n",
    "    # Plot result progress\n",
    "    tsp_plots.plot_result_progress_cross_entropy(\n",
    "        all_results_batch_number, \n",
    "        all_results_average_reward, \n",
    "        all_results_maxumum_reward)\n",
    "    \n",
    "    # Plot best route\n",
    "    co_ords = env.state.city_locations\n",
    "    best_route = [0] + best_route\n",
    "    route_co_ords = [co_ords[i] for i in best_route]\n",
    "    tsp_plots.plot_route(best_route, route_co_ords)\n",
    "    \n",
    "    ###################### Show route and distances ############################\n",
    "    \n",
    "    print ('Route')\n",
    "    print (best_route)\n",
    "    print ()\n",
    "    print ('Best route distance')\n",
    "    print (f'{env.state.calculate_distance(best_route):.0f}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "New best average. Batch :     1,  Time    0  Average reward   -849\n",
      "Batch   2. Average and best run:   -874,    655. \n",
      "New best average. Batch :     3,  Time    0  Average reward   -680\n",
      "Batch   6. Average and best run:   -695,    804. \n",
      "New best average. Batch :     7,  Time    0  Average reward   -616\n",
      "Batch   9. Average and best run:   -738,    541. \n",
      "New best average. Batch :    10,  Time    0  Average reward   -601\n",
      "Batch  12. Average and best run:  -1002,    520. \n",
      "New best average. Batch :    13,  Time    0  Average reward   -529\n",
      "Batch  23. Average and best run:   -724,    642. \n",
      "New best average. Batch :    24,  Time    0  Average reward   -440\n",
      "Batch  36. Average and best run:   -455,    869. \n",
      "New best average. Batch :    37,  Time    1  Average reward   -419\n",
      "Batch  47. Average and best run:   -522,    670. \n",
      "New best average. Batch :    48,  Time    1  Average reward   -417\n",
      "Batch  49. Average and best run:   -503,    387. \n",
      "New best average. Batch :    50,  Time    1  Average reward   -320\n",
      "Batch  53. Average and best run:   -433,    701. \n",
      "New best average. Batch :    54,  Time    1  Average reward   -282\n",
      "Batch  69. Average and best run:   -326,    893. \n",
      "New best average. Batch :    70,  Time    1  Average reward   -264\n",
      "Batch  73. Average and best run:   -300,    709. \n",
      "New best average. Batch :    74,  Time    1  Average reward   -240\n",
      "Batch  80. Average and best run:   -315,    848. \n",
      "New best average. Batch :    81,  Time    1  Average reward   -186\n",
      "Batch  94. Average and best run:   -378,    545. \n",
      "New best average. Batch :    95,  Time    1  Average reward    -85\n",
      "Batch 108. Average and best run:   -154,    823. \n",
      "New best average. Batch :   109,  Time    1  Average reward     44\n",
      "Batch 128. Average and best run:     -2,    783. \n",
      "New best average. Batch :   129,  Time    2  Average reward     87\n",
      "Batch 130. Average and best run:      7,    782. \n",
      "New best average. Batch :   131,  Time    2  Average reward    178\n",
      "Batch 138. Average and best run:    136,    653. \n",
      "New best average. Batch :   139,  Time    2  Average reward    246\n",
      "Batch 142. Average and best run:    181,    907. \n",
      "New best average. Batch :   143,  Time    2  Average reward    263\n",
      "Batch 143. Average and best run:    263,    897. \n",
      "New best average. Batch :   144,  Time    2  Average reward    291\n",
      "Batch 148. Average and best run:    153,    912. \n",
      "New best average. Batch :   149,  Time    2  Average reward    316\n",
      "Batch 160. Average and best run:    310,    865. \n",
      "New best average. Batch :   161,  Time    2  Average reward    342\n",
      "Batch 171. Average and best run:    258,    915. \n",
      "New best average. Batch :   172,  Time    2  Average reward    418\n",
      "Batch 178. Average and best run:    318,    838. \n",
      "New best average. Batch :   179,  Time    2  Average reward    420\n",
      "Batch 179. Average and best run:    420,    959. \n",
      "New best average. Batch :   180,  Time    2  Average reward    433\n",
      "Batch 180. Average and best run:    433,   1017. \n",
      "New best average. Batch :   181,  Time    2  Average reward    474\n",
      "Batch 186. Average and best run:    381,    993. \n",
      "New best average. Batch :   187,  Time    2  Average reward    491\n",
      "Batch 189. Average and best run:    483,    931. \n",
      "New best average. Batch :   190,  Time    2  Average reward    521\n",
      "Batch 197. Average and best run:    487,    994. \n",
      "New best average. Batch :   198,  Time    2  Average reward    531\n",
      "Batch 198. Average and best run:    531,   1061. \n",
      "New best average. Batch :   199,  Time    2  Average reward    580\n",
      "Batch 203. Average and best run:    573,   1056. \n",
      "New best average. Batch :   204,  Time    2  Average reward    611\n",
      "Batch 204. Average and best run:    611,   1020. \n",
      "New best average. Batch :   205,  Time    2  Average reward    616\n",
      "Batch 207. Average and best run:    589,   1113. \n",
      "New best average. Batch :   208,  Time    3  Average reward    634\n",
      "Batch 210. Average and best run:    592,    989. \n",
      "New best average. Batch :   211,  Time    3  Average reward    645\n",
      "Batch 216. Average and best run:    522,   1079. \n",
      "New best average. Batch :   217,  Time    3  Average reward    717\n",
      "Batch 220. Average and best run:    709,   1017. \n",
      "New best average. Batch :   221,  Time    3  Average reward    741\n",
      "Batch 224. Average and best run:    691,   1093. \n",
      "New best average. Batch :   225,  Time    3  Average reward    794\n",
      "Batch 229. Average and best run:    712,   1125. \n",
      "New best average. Batch :   230,  Time    3  Average reward    795\n",
      "Batch 231. Average and best run:    772,   1107. \n",
      "New best average. Batch :   232,  Time    3  Average reward    818\n",
      "Batch 232. Average and best run:    818,   1130. \n",
      "New best average. Batch :   233,  Time    3  Average reward    842\n",
      "Batch 238. Average and best run:    807,   1099. \n",
      "New best average. Batch :   239,  Time    3  Average reward    855\n",
      "Batch 240. Average and best run:    854,   1125. \n",
      "New best average. Batch :   241,  Time    3  Average reward    866\n",
      "Batch 241. Average and best run:    866,   1159. \n",
      "New best average. Batch :   242,  Time    3  Average reward    888\n",
      "Batch 242. Average and best run:    888,   1125. \n",
      "New best average. Batch :   243,  Time    3  Average reward    963\n",
      "Batch 243. Average and best run:    963,   1159. \n",
      "New best average. Batch :   244,  Time    3  Average reward    965\n",
      "Batch 248. Average and best run:    943,   1137. \n",
      "New best average. Batch :   249,  Time    3  Average reward    972\n",
      "Batch 250. Average and best run:    953,   1144. \n",
      "New best average. Batch :   251,  Time    3  Average reward   1001\n",
      "Batch 251. Average and best run:   1001,   1144. \n",
      "New best average. Batch :   252,  Time    3  Average reward   1025\n",
      "Batch 253. Average and best run:    996,   1142. \n",
      "New best average. Batch :   254,  Time    3  Average reward   1056\n",
      "Batch 262. Average and best run:    540,   1159. \n",
      "New best average. Batch :   263,  Time    3  Average reward   1101\n",
      "Batch 263. Average and best run:   1101,   1159. \n",
      "New best average. Batch :   264,  Time    3  Average reward   1103\n",
      "Batch 265. Average and best run:    851,   1144. \n",
      "New best average. Batch :   266,  Time    3  Average reward   1131\n",
      "Batch 269. Average and best run:   1000,   1159. \n",
      "New best average. Batch :   270,  Time    3  Average reward   1138\n",
      "Batch 272. Average and best run:   1132,   1143. \n",
      "New best average. Batch :   273,  Time    3  Average reward   1140\n",
      "Batch 277. Average and best run:   1135,   1143. \n",
      "New best average. Batch :   278,  Time    3  Average reward   1142\n",
      "Batch 281. Average and best run:   1138,   1143. \n",
      "New best average. Batch :   282,  Time    3  Average reward   1142\n",
      "Batch 293. Average and best run:   1122,   1143. \n",
      "New best average. Batch :   294,  Time    3  Average reward   1143\n",
      "Batch 294. Average and best run:   1143,   1143. "
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Route\n",
      "[0, 5, 9, 4, 7, 6, 1, 2, 3, 8, 0]\n",
      "\n",
      "Best route distance\n",
      "255\n"
     ]
    }
   ],
   "source": [
    "main()"
   ]
  }
 ],
 "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}