-
Neural networks are the base on which modern machine learning models such as LLMs, RNNs, and CNNs are built.
-
This project is an attempt to understand and gain intuition about the mathematics behind it and how it translates to an actual code implementation.
-
The Neural_Network directory contains an implementation of a generalized multi-layer neural network that can be tuned to use different numbers of layers, activation functions, and learning rates, with the only dependency being numpy for linear algebra operations. Along with the code is a test dataset and a report of the performance of different activation functions.
-
After gaining an understanding of NNs, I implemented a convolutional neural network using a higher level of abstraction and convenience by leveraging PyTorch. This can be found in the CNN directory along with a report of the different architectures experimented with.