smokja / self-written-ai Goto Github PK

I think there are some fundamental mistakes in the forward propagation process of my neural network. By the time writing the code I may have not known how to calculate the value of a node and using the activation function properly.

There was an older implementation that did not follow any specific rules and just failed fundamentally. While training the function I am trying to implement it with a reinforcement strategy and applying the new weights with Backpropagation.