Machine Learning
Research For Science
Welcome! My name is Tim Whitaker and Wits End is my personal website, a repository for my neural network research, and an outlet for digging deeper into artificial intelligence.
Data Augmentation
A strategy used in machine learning to increase the diversity and amount of training data without actually collecting new data. It involves creating modified versions of existing data using techniques like rotation, scaling, flipping, cropping, and brightness or color adjustments.
Read MoreQuantization
Quantization is used to reduce the precision of the weights and biases in a model in order to decrease computational requirements. It involves converting full-precision 32-bit weights into lower-precision formats. Typically 16-bit or 8-bit quantization is used, but research has shown promise in resource constrained enviroments for ternary and binary networks.
Read MoreSubnetwork Ensembles
My PhD research explores this idea of breaking up trained deep neural networks into multiple sparse subnetworks that can be trained, modified, and evaluated independently. Generating ensembles in this manner is an incredibly efficient way to improve the generalization performance and parametric utilization of large networks while minimizing computational cost. This work has a natural connection to the topological structures we find in biological brains and it's a beautiful framework for exploring network optimization. I'm currently working on my dissertation, but I'm very excited to share what I have so far...
Read MoreGet In Touch
Hero images generated with neural networks via midjourney.