Mo T
- Research Program Mentor
PhD Doctor of Philosophy candidate
Expertise
Artificial intelligence, machine learning, statistical learning, deep learning
Project ideas
Optimizing a state-of-the-art k-medoids algorithm in C++
We've developed a state-of-the-art clustering algorithm for the k-medoids problem (similar to k-means, but where the cluster centers must be actual datapoints). Our work was accepted to Neural Information Processing Systems (NeurIPS) 2020. We have also released a Python package, written in C++, to enable others to use our algorithm. We are looking for someone to optimize this package. Experience with multithreading, caches, and profiling code is a plus.
Learning the best data augmentations
Data augmentation in machine learning is currently a bag of tricks. Very little work exists regarding what is the "best" form of data augmentation; most people try everything and keep what sticks. Our work aims to learn the best data augmentations from the training data itself. We have promising initial results and are looking for someone who can help contribute to the research. We expect this research will result in a publication. To see if you'd be a good fit for this project, you should test your understanding of the AutoAugment paper. Experience with Bayesian Optimization or reinforcement learning is a plus.