I primarily study tensor networks, and their uses in machine learning and physics.
Supervised Program for Alignment Research
I work on capturing temporal feature correlation in LLMs with Dmitry Manning-Coe.
→ wip
Lab for Parallel Numerical Algorithms
I work on tensor algorithms with Edgar Solomonik.
→ see active projects
I. Bayesian Tensor Decompositions
We are exploring Bayesian algorithms for tensor decomposition and probabilistic linear algebra. Currently, we are building on Alternating Mahalanobis Distance Minimization (AMDM) to design a decomposition method with mode-wise covariance modeling.
More updates soon!
II. MCMC for Tensor Network Contractions
We are developing a Markov chain Monte Carlo algorithm to estimate contractions of closed tensor networks. This allows efficient approximation of contraction quantities using methods from statistical physics to improve mixing. Such contractions \[ \text{Tr}(ABCD) = \sum_{ijkl} A_{ij}B_{jk}C_{kl}D_{li} \]are trivial in small cases but become \(\#\textsf{P}\)-hard for large networks. Our approach targets use cases in quantum circuits and chemistry, where exact contraction is intractable.