Info

Project which allow to simulate simple biological neural networks.

Below I attach the screen of a sample simulation. That simulation explains how lateral inhibition works.

Installation

1. Download tar.gz file.
2. Run pip install ./path_to_file/cereblus-0.1.0.tar.gz

Usage

We have got four basic objects

1. Neuron represents one neuron.
2. Neural Network represents network.
3. Displayer represents graphical interface.
4. Stimulus represents stimulus, which activates neurons.

Creating first network

Let's write

from cereblus import *

neural_network = NeuralNetwork()
n1 = neural_network.neuron(template=Receptor, pos=(300, 300, 0))
n2 = neural_network.neuron(template=Receptor, pos=(300, 400, 0))
n3 = neural_network.neuron(template=Receptor, pos=(550, 350, 0))

We've just created a network with three neurons – n1, n2 and n3. First two neurons are receptors -- they can react to the stimuli, which we will use later. n3 is standard Neuron. It is worth mentioning that all Receptor in child class of Neuron. Let's join them

n1.connect(n3, coeff=0.5)
n2.connect(n3, coeff=0.7)

Now n1 is connected to n3 with coefficient 0.5 and n2 is connected to n3 with coefficient 0.7.

We connect our network to the Displayer

displayer = Displayer(neu_net)
displayer.show()

Stimulating neurons

Our network is doing exactly nothing. Why? We haven't stimulated it. Let's create file stimulation.json

  {
  "num_of_pixels": 2,
  "loop": true,
  "phases": [
    [0.5, "O-"],
    [0.5, "-O"],
    [1, "OO"]
  ]
}

And connect it to the neurons n1 and n2

stimulus = StimulationLoader.load_from_file("stimulation.json")
stimulus.connect(n1, num_of_pixel=0)
stimulus.connect(n2, num_of_pixel=1)

We can imagine it as two receptors connected to the two pixels. Pixel is active when there is 'O' in current phase, and it in inactive when it is '-'.

Now let's look on our network.

Steps further

You can write your own Neurons – which for example learn and change connections. The easiest way is to inherit from the Neuron class. Look into the code to undersand how Neuron works – it is simple.