Skip to content

felix-andreas/hopfieldnetwork

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

hopfieldnetwork

A Hopfield network is a special kind of an artifical neural network. It implements a so called associative or content addressable memory. This means that memory contents are not reached via a memory address, but that the network responses to an input pattern with that stored pattern which has the highest similarity.

hopfieldnetwork is a Python package which provides an implementation of a Hopfield network. The package also includes a graphical user interface.

Installing

Install and update using pip:

pip install -U hopfieldnetwork

Requirements

  • Python 2.7 or higher (CPython or PyPy)
  • NumPy
  • Matplotlib

Usage

Import the HopfieldNetwork class:

from hopfieldnetwork import HopfieldNetwork

Create a new Hopfield network of size N = 100:

hopfield_network1 = HopfieldNetwork(N=100)

Save / Train Images into the Hopfield network:

hopfield_network1.train_pattern(input_pattern)

Start an asynchronous update with 5 iterations:

hopfield_network1.update_neurons(iterations=5, mode="async")

Compute the energy function of a pattern:

hopfield_network1.compute_energy(input_pattern)

Save a network as a file:

hopfield_network1.save_network("path/to/file")

Open an already trained Hopfield network:

hopfield_network2 = HopfieldNetwork(filepath="network2.npz")

Graphical user interface

Hopfield network GUI

In the Hopfield network GUI, the one-dimensional vectors of the neuron states are visualized as a two-dimensional binary image. The user has the option to load different pictures/patterns into network and then start an asynchronous or synchronous update with or without finite temperatures. There are also prestored different networks in the examples tab.

Start the UI:

If you installed the hopfieldnetwork package via pip, you can start the UI with:

hopfieldnetwork-ui

Otherwise you can start UI by running gui.py as module:

python -m hopfieldnetwork.gui

GUI Layout

The Hopfield network GUI is divided into three frames:

Input frame
The input frame (left) is the main point of interaction with the network. The user can change the state of an input neuron by a left click to +1, accordingly by to right-click to -1. This will only change the state of the input pattern not the state of the actual network. The input pattern can be transfered to the network with the buttons below:

  • Set intial sets the current input pattern as the start configuration of the neurons.
  • Save / Train stores / trains the current input pattern into the Hopfield network.
  • Rand sets a random input pattern.
  • Clear sets all points of the input pattern to -1.

Output frame
The output frame (center) shows the current neuron configuration.

  • Sync update starts a synchronous update.
  • Async update starts an asynchronous update.
  • Randomize randomly flips the state of one tenth of the neurons.
  • Set partial sets the first half of the neurons to -1.
  • Set random sets a random neuron state.

Saved pattern frame
The Saved pattern frame (right) shows the pattern currently saved in the network.

  • Set initial sets the currently displayed image as new neuron state.
  • Set input sets the currently displayed image as input pattern.
  • Remove removes the currently displayed image from the Hopfield network.

Menu bar

  • In the Network tab, a new Hopfield network of any size can be initialized.

In addition, it is possible to save the current network and load stored networks. Also, a raster graphic (JPG, PNG, GIF, TIF) can be added to the network or an entirly new network can be created out of multiple images.

  • In the Options tab, the update with finite temperatures can be (de)activated.
  • View offers options for visually changing the GUI.
  • In the Examples tab, different example networks can be loaded.

License

GNU General Public License v3.0