pycustomnets 0.0.3

A Neural Network customization tool

pycustomnets

This is a little project that I've been working on for a while with the primary goal of gaining a more in depth understanding of deep learning and honing various skills

This is meant to be my version of tensorflow and while its definitely not as good as tensorflow for now, It does have same basic functions.

This project is also a package which can be downloaded through pip (It doesn't work yet, will be fixed soon maybe)

Activation functions

• Leaky Relu ("lrelu")
• Relu ("relu")
• Sigmoid ("sigmoid")
• Softmax ("softmax" but this is not yet implemented in the class)

Error Functions

• Mean Squared Error (cannot be modified for now)

Optimization algorithms

• Stochastic Gradient Descent ("sgd")
• ADAM ("adam")

In the future, I hope to implement Softmax properly, add Cross Entropy and Categorical Cross Entropy, add Reinforcement Learning algorithms and other stuff. I will also be removing the example code from the def file.