CSE 190: Deep Learning

Programming Assignment 3, Fall 2018

Learning Objectives

1. Understand the basics of convolutional neural networks, including convolutional layer mechanics, max-pooling, and learned parameters/feature maps.
2. Learn how to implement a CNN architecture in PyTorch for image/object classification using best practices.
3. Build intuition on the effects of modulating the design of a CNN by experimenting with your own (or "classic") architectures.
4. Learn how to address the imbalanced/rare class prediction problem in multiclass classification.
5. Visualize and interpret learned feature maps of a CNN.