This codebase provides the skeletal structure for implementing neural networks from scratch, exclusively in numpy. Fill in the blanks to implement three fundamental neural network architectures: feedforward, recurrent, and convolutional. See instructions.pdf for a walkthrough of how to complete the repo and test your models!
I wrote this codebase while a Graduate Student Instructor for UC Berkeley's Machine Learning course, CS189/289A. It was used as the 6th homework assignment in the Spring 2020 semester. sagnibak is a contributor.