projects
Final talk for PHY562 (graduate biophysics).
A method for splitting ML datasets that maximizes the dissimilarity between splits while maintaining a target split ratio.
Recovering the phase transition and order parameter of the 2D Ising model in machine-learned latent manifolds.
Does automated science actually work? Thinking about this question got me interested in information efficiency as a metric for scientific automation.
How will AI performance scaling occur for domains outside next-token prediction for human language? Work done at Two Sigma.
Interactive visualization of biological processes across length and time scales.
Build and simulate gene regulatory networks with activation and repression dynamics.
ASCII visualization of Ising model dynamics.