Jenny Chen, Linnea May
Neural networks, deep learning, artificial intelligence... Everyone seems to know these buzzwords, but very few truly understand how they work. The goal of our project is to try and change that just a little, by building tools to visualize how a neural net is training.
This project had two main goals:
- As a team new to deep learning, we want to learn more about neural networks, from typical ANNs to CNNs.
- In order to help other people gain the same understanding, build tools to help show how these netowrks function.
Our approach overall has been to look at the evolution of learned representations of a neural net by projecting high-dimensional vectors of activations for a certain input at a given layer into 2D space using a variety of dimensionality reduction techniques.