General Theory for Group Resetting with Application to Avoidance

Juhee Lee, Seong-Gyu Yang, Hye Jin Park, Ludvig Lizana · 2024

Read General Theory for Group Resetting with Application to Avoidance by Juhee Lee, Seong-Gyu Yang, Hye Jin Park, Ludvig Lizana (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-02524.txt.

Introduction

Juhee Lee and co-authors build a general theory in 2024 for a process where a whole group restarts, not just one member. In the classic setup, a lone particle wanders and is now and then sent back to a start point. Here, the restart is driven by the group as a whole, for instance by its most extreme member.

Settings like this show up in bacteria facing antibiotics and in robot swarms that search an area. The team treats the group's center as a single stand in particle and writes an equation for how it spreads. From that they get exact formulas for where the center settles and how widely it spreads. They also study a case where the group must steer clear of a bad region. The theory gives a shared base for group restart problems that were solved one at a time before.

Catalog record

CREATORJuhee Lee, Seong-Gyu Yang, Hye Jin Park, Ludvig Lizana
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE22 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-02524

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

General Theory for Group Resetting with Application to Avoidance Juhee Lee, Seong-Gyu Yang, Hye Jin Park, Ludvig Lizana Abstract We present a general theoretical framework for group resetting dynamics in multi-particle systems in a drift potential. While traditional resetting models typically focus on a single particle, our setup considers a group of particles whose collective dynamics determine the resetting process. More recently, resetting has also been used as a regulatory mechanism to avoid adverse outcomes, such as preventing critically high water levels in dams or deleveraging financial portfolios. Here, we extend current resetting theories to group dynamics, with applications ranging from bacterial evolution under antibiotic pressure to multiple-searcher optimization algorithms. Combining renewal theory and extreme value statistics, we derive a Fokker-Planck equation for the spatial distribution of the group’s center of mass, treated as an effective particle. This formalism allows us to analytically calculate essential observables including the stationary mean position, variance, and a dimensionless measure of risk–the squared coefficient of variation. Our results demonstrate how the group size and resetting rate affect the probability of avoiding the danger region. This theoretical approach opens new perspectives on designing optimal group-level search and avoidance strategies through resetting. In traditional search with stochastic resetting, researchers aim to derive the optimal resetting rate that minimizes the time to find a designated target Evans and Majumdar 2011 ; Pal and Reuveni 2017 ; Dahlenburg et al. 2021 ; Chechkin and Sokolov 2018 ; Durang et al. 2019 . While initially formulated as memoryless, the resetting theory now incorporates non-Markovian processes Harbola 2024 ; Shkilev 2017 ; Shkilev and Sokolov 2022 , nonequilibrium environments Masó-Puigdellosas et al. 2019 ; Goswami 2025 , and bounded or nonrenewal systems Bodrova et al. 2019 ; Méndez et al. 2022 . These developments have proven useful in several applications, from transport processes to optimal search strategies. Yet, there is a broad class of problems that fall outside this single-particle framework, such as systems with many searchers where the resetting depends on collective behaviors. One example is swarm-search in optimization algorithms Mesquita et al. 2008 ; Kennedy and Eberhart 1995 ; Wang et al. 2018 . Here, multiple random walkers explore a mathematical landscape in parallel. During the search, agents continuously exchange “fitness” values (quantified with respect to an objective function) and periodically reset their positions to that of the fittest member. While this strategy can accelerate convergence, it raises nontrivial questions about the optimal number of searchers and the resetting rate. Another example is bacterial evolution under antibiotic pressure. When under such attack, the bacterial population drifts toward a drug-resistant “super-bacteria” state by evolving their metabolisms Davies and Davies 2010 ; Neu 1992 ; Munita and Arias 2016 ; Blair et al. 2015 . This represents an example of group search in biological-trait space, which could be halted by constantly resetting the population to the least fit bacteria using an artificial selection protocol Arias-Sánchez et al. 2019 ; Sánchez et al. 2021 ; Thomas et al. 2024 . Just like in the optimization problem, finding the relevant parameters in such a scheme, such as the best resetting rate, represents a considerable challenge. Inspired by these examples, we develop a general framework for group resetting in a drift potential, where the system’s collective dynamic state determines the resetting point. Traditional models often use simple rules for resetting, such as returning to the starting point or a fraction of the traveled distance Dahlenburg et al. 2021 . In contrast, we incorporate group-level dynamics, where resetting targets the particle that has traveled the farthest. Our framework rests on renewal theory with extreme value statistics, from which we derive a master equation for the spatial distribution of an effective particle representing the system’s center-of-mass motion (CM). To further demonstrate the theory, we apply it to a group of particles diffusing in a harmonic potential that pulls them toward a dangerous region. We solve the corresponding group avoidance problem by computing the probability that the ensemble remains on the safe side. While avoidance has been previously studied in a single-particle setting De Bruyne et al. 2020 ; De Bruyne et al. 2021 (e.g., preventing dam overflow or limiting excessive financial leverage), our work extends this concept to groups, where avoidance is regulated through resetting to the best-performing member. We schematically illustrate the group resetting process and its application to avoidance problem in Figure 1 . In Fig. 1 (a), we depict how all particles relocate simultaneously to the position of a selected particle in the group at the reset, and Fig. 1 (b) provides an example of a particle diffusing in a drift potential, where the left region (red) represent an undesired or dangerous area. In Fig. 1 (c), we show the position distribution before and after the reset, when all particles relocate to one located farthest to the right. Figure 1: Schematic figure of group resetting. (a) When the group resets, all the particles relocate to the position of a certain particle in the group. (b) As an example, each particle diffuses under a drift potential that drags them towards the minimum. The left red side of the potential is an undesired or dangerous region to avoid. (c) The position distributions before and after the group resetting event. Particles reset their positions to the rightmost particle and restart dynamics. Black lines are probability density functions at each step. We consider n independent overdamped Brownian particles with coordinates x_{i} ( i=1,\ldots,n ) in a one-dimensional potential V(x_{i}) . All particles start from x_{i}=x_{0} and diffuse with diffusion constant D . With a resetting rate r , the entire group simultaneously relocates to a position X(t) , generally depending on the particle positions x_{1},\ldots,x_{n} . To describe the collective dynamics, we introduce an effective particle representing the CM coordinate \zeta . It starts at \zeta_{0}=x_{0} , evolves with diffusion constant D/n in an effective potential V_{\text{eff}}(\zeta) (in general, V(x_{i})\neq V_{\text{eff}}(\zeta) ), and resets to \zeta=X(t) with rate r . Following the approach in Ref. Evans and Majumdar 2011 , we describe the spatial distribution of \zeta , \ P(\zeta,t|\zeta_{0},t_{0})\equiv P(\zeta,t) , by the Fokker-Planck equation \begin{split}\frac{\partial}{\partial t}P(\zeta,t)&=\frac{\partial}{\partial\zeta}\left[A(\zeta)P(\zeta,t)\right]+\frac{\partial^{2}}{\partial\zeta^{2}}\left[\frac{D}{n}P(\zeta,t)\right]\\ &-rP(\zeta,t)+rP_{r}(\zeta,t),\end{split} (1) where, the first two terms represent drift and diffusion. The last two terms describe the resetting process to the removal and re-introduction at positions sampled from the resetting probability density P_{r}(\zeta,t) . Calculating P_{r}(\zeta,t) is the main challenge in this problem, as it depends on the process itself. Specifically, it depends on the conditional probability density function (or kernel) K_{n}(\zeta=X|\zeta^{\prime}=X^{\prime};\tau) describing the likelihood of finding the effective particle at \zeta after time \tau since the last reset to \zeta^{\prime} . As in Ref. Evans et al. 2020 , we us renewal theory to relate this kernel to P_{r}(\zeta,t) , yielding \begin{split}P_{r}&(\zeta,t)=\Psi(t)K_{n}(\zeta|\zeta_{0};t)\\ &+\int_{0}^{t}d\tau\,\phi(\tau)\int_{-\infty}^{\infty}d\zeta^{\prime}\,K_{n}(\zeta|\zeta^{\prime};\tau)P_{r}(\zeta^{\prime},t-\tau).\end{split} (2) Here, \phi(\tau) denotes the waiting-time distribution between consecutive resetting events, \Psi(t)=1-\int_{0}^{t}\phi(\tau)d\tau is the survival probability, and the resetting rate is given by r=\left[\int_{0}^{\infty}\tau\phi(\tau)d\tau\right]^{-1} . Together, Eqs. ( 1 ) and ( 2 ), with the initial condition P_{r}(\zeta,0)=\delta(\zeta-\zeta_{0}) , form the core formula for analyzing group resetting dynamics in various scenarios. To proceed further, one must choose a specific resetting scheme, leading to different kernels K_{n}(\zeta|\zeta^{\prime};\tau) . The simplest example is always returning

Read the complete text