Gilad Turok

NYC, US

Hey ! I’m currently a Research Analyst at the Flatiron Institute’s Center for Computational Mathematics working with Bob Carpenter.

My research is focused on (generative, efficient, robust) deep learning and Bayesian statistics (variational inference + Markov chain Monte Carlo).

I like to hack on LLMs, open too many ArXiv tabs, and write open source code. I also train calisthenics and explore NYC.

I graduated from Columbia University with a B.A. in Applied Mathematics, though lowkey I mostly studied machine learning.

At Columbia I did research under Itsik Pe’er on reconstructing evolutionary trees of CRISPR-edited cells with hyperbolic optimization. I also did research under David Knowles on determining which genetic variants cause phenotypic traits with variational inference.

I love when people reach out – so please do!

news

Jul 1, 2024	Attending ISBA 2024 in Venice, Italy and presenting a poster.
May 9, 2023	Interning at the Flatiron Institute with Chirag Modi and Bob Carpenter.

latest posts

Aug 16, 2023	What is Academic Research and Why Do it

selected publications

2024

Sampling from multiscale densities with delayed rejection generalized Hamiltonian Monte Carlo

Gilad Turok, Chirag Modi, and Bob Carpenter

2024

Abs arXiv Code

With the increasing prevalence of probabilistic programming languages, Hamiltonian Monte Carlo (HMC) has become the mainstay of applied Bayesian inference. However HMC still struggles to sample from densities with multiscale geometry: a large step size is needed to efficiently explore low curvature regions while a small step size is needed to accurately explore high curvature regions. We introduce the delayed rejection generalized HMC (DR-G-HMC) sampler that overcomes this challenge by employing dynamic step size selection, inspired by differential equation solvers. In a single sampling iteration, DR-G-HMC sequentially makes proposals with geometrically decreasing step sizes if necessary. This simulates Hamiltonian dynamics with increasing fidelity that, in high curvature regions, generates proposals with a higher chance of acceptance. DR-G-HMC also makes generalized HMC competitive by decreasing the number of rejections which otherwise cause inefficient backtracking and prevents directed movement. We present experiments to demonstrate that DR-G-HMC (1) correctly samples from multiscale densities, (2) makes generalized HMC methods competitive with the state of the art No-U-Turn sampler, and (3) is robust to tuning parameters.