Christopher Tosh

I am an associate research scientist in Computational Oncology at Memorial Sloan Kettering Cancer Center in the lab of Wesley Tansey. My current interests are on problems arising in interactive learning, representation learning, and robust learning, particularly with applications to cancer research.

Before MSKCC, I was a postdoc at Columbia University in the Data Science Institute, supervised by Daniel Hsu. Before that, I received my PhD in Computer Science from UC San Diego, where I was advised by Sanjoy Dasgupta.

selected publications

Simple and near-optimal algorithms for hidden stratification and multi-group learning

C. Tosh, and D. Hsu

In Proceedings of the 39th International Conference on Machine Learning, 2022

Abs arXiv HTML PDF Errata

Multi-group agnostic learning is a formal learning criterion that is concerned with the conditional risks of predictors within subgroups of a population. The criterion addresses recent practical concerns such as subgroup fairness and hidden stratification. This paper studies the structure of solutions to the multi-group learning problem, and provides simple and near-optimal algorithms for the learning problem.
Contrastive learning, multi-view redundancy, and linear models

C. Tosh, A. Krishnamurthy, and D. Hsu

In Proceedings of the 32nd International Conference on Algorithmic Learning Theory, 2021

Abs arXiv HTML PDF

Self-supervised learning is an empirically successful approach to unsupervised learning based on creating artificial supervised learning problems. A popular self-supervised approach to representation learning is contrastive learning, which leverages naturally occurring pairs of similar and dissimilar data points, or multiple views of the same data. This work provides a theoretical analysis of contrastive learning in the multi-view setting, where two views of each datum are available. The main result is that linear functions of the learned representations are nearly optimal on downstream prediction tasks whenever the two views provide redundant information about the label.
Diameter-based active learning

C. Tosh, and S. Dasgupta

In Proceedings of the 34th International Conference on Machine Learning, 2017

Abs arXiv HTML PDF

To date, the tightest upper and lower-bounds for the active learning of general concept classes have been in terms of a parameter of the learning problem called the splitting index. We provide, for the first time, an efficient algorithm that is able to realize this upper bound, and we empirically demonstrate its good performance.