Jonathan Weare
New York University, NY·Mathematics & Statistics
Some of the most consequential events in nature are also the rarest: a protein complex coming apart, or a once-in-a-century heat wave. Rare events are hard to predict precisely because they almost never show up in the data we can generate or observe.
- Fields
Mathematics & Statistics
Computer Science
Earth & Environmental Science
- Topics
rare event sampling
molecular dynamics
mean first-passage time
- Grants
- 2 known grants, through May 2030
What they work on
- Develop schemes that estimate dynamical statistics such as mean first-passage times and committors from datasets of short molecular dynamics trajectories, avoiding the need for trajectories long enough to contain the rare event.
- Generate the required short-trajectory datasets either with rare-event sampling methods previously developed by the team or by seeding initial conditions throughout a molecular system's state space.
- Apply the approach to slow biomolecular processes such as dissociation of the insulin dimer, the step required before insulin binds its receptor, and to voltage-sensing proteins.
- Build a strategy for tuning simulation models so their computed rates and related quantities better match experimentally obtained estimates.
- Develop AI Dynamic Galerkin Approximation to extract long return periods for extreme climate events from large ensembles of short AI emulator runs.
- Develop AI-RES, an iterative scheme combining rare-event sampling with numerical solvers and AI emulators, to estimate return periods for heat waves and cold snaps and generate new rare-event data for retraining the emulator.
More about the research
Simulations can resolve the fast motions of molecules, and climate models the day-to-day weather, but waiting for the rare event itself can take astronomically more computer time than is available, and machine-learning models trained on typical data have almost no examples of extremes to learn from. The common thread across these projects is mathematics for estimating how often rare events happen, and how they unfold, from large collections of short simulations rather than a few impossibly long ones.
Why it matters
Heat waves and cold spells cause major societal harm, and more accurate estimates of how often extreme events of a given severity recur would give planners and climate scientists better numbers to work with, including for events no model has seen before. On the molecular side, being able to compute rates for slow processes such as insulin dimer dissociation or the motion of voltage-sensing proteins could reveal physical mechanisms relevant to signaling and drug action in far more detail than direct simulation allows.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Dorian Abbot
University of Chicago, IL·Earth & Environmental Science
Develop AI Dynamic Galerkin Approximation (AI-DGA) to estimate long return periods, the typical time between extreme events, for heat waves and cold snaps from large ensembles of short AI emulator runs.
Lantao Sun
Colorado State University-Fort Collins, CO·Earth & Environmental Science
Determine how much Arctic sea ice loss affects summer atmospheric circulation and the jet stream, and through what mechanisms, using DAMIP and PAMIP climate model simulations.
Jiakai Zhang
New Mexico Institute of Mining and Technology, NM·Earth & Environmental Science
Integrate large public spatial, hazard and socioeconomic datasets into a single AI-driven model of regional environmental risk and vulnerability across the United States.
The school
New York University: 72 labs on PI Pages, $1.5B a year on research, 540 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.