Homework 1:
Goal: Focuses on foundational machine learning and neural network concepts, including implementing perceptron, logistic regression, and multi-layer perceptrons from scratch, followed by experiments with PyTorch-based feedforward networks and analysis of model capacity (width/depth). It also explores alternative activation functions such as sparsemax and relumax, including their mathematical properties and optimization behavior.
Grade: 94/100
Homework 2:
Goal: Centers on deep learning applications using PyTorch, starting with convolutional neural networks for image classification and extending to sequence-based regression for biological data (RNA–protein binding prediction). It involves designing and comparing neural architectures (e.g., CNNs, RNNs), incorporating attention mechanisms, and evaluating performance using appropriate metrics and validation strategies.
Grade: 100/100