This repository contains the deep learning components implemented from scratch and with pytorch. For in detailed theory of deep learning, mathemetics behind it's operations I referred the Dive into Deep Learning book. The code part in this book is based on mxnet. For the sake of practice I tried to implemnted all the components into pytorch / numpy or useing basic python functionalities.
The d2l book is already porte into Pytorch based codes and can be found here. My implementation of all the functions that form components of a deep neural network can be found in dl_functions.py script.
- Linear Neural Network: Linear and Softmax regression implemented from scratch and with Pytorch
- MLP: Implemented a concept of Multi Layer Perceptron, l2 regularization and dropout.
- Deep learning Computations: Conceptualized layers, blocks and custom layers
- Convolutional Neural Network: Implemented Conv layer, pooling layer, batch normalization layer from scratch