/Crude_NeuralNet

A simple neural network only with numpy, no other packages (such as Keras, Tensorflow, etc)

Primary LanguagePython

Pure Neural Network in Python

I wrote this NN so I could experience the network functionality directly. This means I did not use any usual library for NNs in Python, such as Keras or Tensorflow. I did, however, use Numpy, which I believe does not defeat the purpose of this task.

Below the functions you should see that there are two tasks available (both commented).

First: A random network and plots

First generates a random net, with number of layer equal to entries in NS, and layer size is an int. You can plot it with the Plot_2Dnetwork2()function.

Second: an OR logic gate using a 2 layer NN

Just follow the written steps to get the following plot:

Plot 1: OR gate with 2 layer network

Some resulting plots

It is interesting to see that the network gets better at reproducing the randomness of the inputs as more layers are added, but also smooths them more.

Plot 2: 6 layer random Net

Plot 3: 9 layer random Net

Plot 4: 14 layer random Net

Plot 5: 22 layer random Net