We introduce a general framework for adaptive resetting, where stochastic resetting depends on the state and age of the process. Using a reweighting scheme applied to trajectories without resetting, we efficiently compute key observables like first-passage times and steady-states. This enables the design of informed search strategies and complex non-equilibrium behaviors without brute-force sampling. We further develop a machine learning approach to optimize adaptive resetting protocols and demonstrate its effectiveness in accelerating molecular dynamics simulations.

Congratulations to Tommer David Keidar for winning the David and Paulina Trotsky Foundation Fellowship for the school year 2024-25.

We introduce informed stochastic resetting, a novel strategy for enhancing molecular dynamics simulations by selectively resetting only when the reaction coordinate indicates insufficient progress. Applied to a model system and chignolin in explicit water, this approach accelerates Metadynamics simulations by 2–3 orders of magnitude. By leveraging information about reaction progress, informed resetting significantly extends the applicability of stochastic resetting for overcoming time scale limitations in molecular simulations.

Machine learning models have become indispensable tools in applications across the physical sciences. However, their training is often time-consuming. Several protocols have been developed to perturb the learning process and improve the training. However, their design is usually done ad hoc by intuition and trial and error. To rationally optimize training protocols, we frame them as first-passage processes. This reveals that a model’s response at a single perturbation frequency can predict its behavior across others.

We derive high-order Michaelis-Menten equations that extend the classical linear relation between mean turnover time and reciprocal substrate concentration. These equations reveal universal linear relations for higher-order moments, granting access to previously inaccessible kinetic parameters. We show how key observables—such as the enzyme-substrate lifetime, the binding rate, and the probability of successful catalysis—can all be inferred from these relations.

We introduce smart resetting—a strategy that leverages information to selectively reset a diffusing particle only when it benefits its progress toward a target. Our analysis demonstrates that smart resetting consistently lowers the energy cost per mean first passage time compared to regular resetting, achieving the known minimum energy cost bound for a diffusing particle regardless of the resetting rate. We extend our findings to consider nonlinear energetic cost functions and drift-diffusion processes, offering deeper insights into the interplay between information and resetting.

Congratulations to Itamar Shitrit for winning the outstanding student's scientific achievement award of the center for Physics and Chemistry of living systems at Tel Aviv University. The prize was awarded for his research – Sokoban percolation on the Bethe lattice – which was carried out in collaboration with Ofek Lauber Bonomo.

Congratulations to Tommer David Keidar for winning the outstanding poster award as part of the 88th annual meeting of the Israel Chemical Society. The award was given for his research on the universal linear response of the mean first-passage time.

We develop a comprehensive theory of queueing systems with service resetting, where service times depend on both the server's intrinsic slowdown and the job's inherent size. Analyzing both Poissonian and deterministic resetting policies, we derive conditions under which resetting reduces the mean service time to improve queue performance. Our results, validated through numerical simulations, highlight service resetting as an effective tool for mitigating service time fluctuations in diverse queueing scenarios.