I took a deep learning class in college and these were some of the hws.
(This was written by OG ChatGPT) This repository contains separate implementations of neural network models using PyTorch. Each implementation focuses on a different topic or task. The following is a brief description of each tutorial:
This tutorial implements a Generative Adversarial Network (GAN) using PyTorch to generate realistic-looking images of handwritten digits from the MNIST dataset. The GAN architecture consists of a generator and a discriminator network trained in an adversarial manner.
In this tutorial, a self-attention mechanism and the Vision Transformer (ViT) architecture are implemented using PyTorch. The ViT model applies self-attention to images, achieving state-of-the-art results in various vision tasks. The tutorial explores the ViT architecture and its application to image recognition.
This tutorial focuses on self-supervised contrastive learning, a technique used for unsupervised learning without labels. The implementation uses PyTorch to train a model using contrastive learning to cluster images and their augmented versions in a latent space. The SimCLR method is introduced and implemented as an example of contrastive learning.
In this assignment, an image captioning model is implemented using convolutional neural networks (CNNs) and recurrent neural networks (RNNs), specifically Long Short-Term Memory (LSTM), with PyTorch. The model takes images as input and generates natural language descriptions for them. The Microsoft COCO dataset is used for training and evaluation.
This homework assignment focuses on implementing a simple neural network using NumPy. The goal is to gain practice with Python and NumPy while building intuition about the forward and backward propagation algorithms. The implementation follows a basic neural network architecture and includes parameter initialization, cost function calculation, and gradient descent optimization.