CNN in plain numpy

Implementation of Convolutional neural network using only NumPy library.

---

This project was created to better understand the CNN blocks, their forward and backward pass. The idea was taken from practical assignments to Dr. Andrew Ng's Convolutional Neural Nets course.

Project includes following implementations:

• Convolution2D, Maxpooling2D, Dense, Flatten layers
• Binary and Categorical Cross Entropy functions
• ReLU and Sigmoid activation functions
• Mini Batch stochastic gradient descent optimizer

Table of contents

• Main blocks
• Activation functions
• Loss functions
• Optimizers
• TODO

Main blocks

Activation functions

Loss functions

Optimizers

TODO

1. Finish documentation
2. Finish Readme