A Python implementation of the Hopfield network
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.
Install and update using pip:
pip install -U hopfieldnetwork
- Python 2.7 or higher (CPython or PyPy)
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:
Start an asynchronous update with 5 iterations:
Compute the energy function of a pattern:
Save a network as a file:
Open an already trained Hopfield network:
hopfield_network2 = HopfieldNetwork(filepath="network2.npz")
Graphical user interface
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:
Otherwise you can start UI by running
gui.py as module:
python -m hopfieldnetwork.gui
The Hopfield network GUI is divided into three frames:
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.
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.
- 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.
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size hopfieldnetwork-1.0.1-py3-none-any.whl (88.6 kB)||File type Wheel||Python version py3||Upload date||Hashes View|
Hashes for hopfieldnetwork-1.0.1-py3-none-any.whl