应伟德体育伟德体育官方网站邀请,北马其顿科学与艺术公司Petar Jolakoski博士于2025年11月1日至11月5日访问伟德体育伟德体育官方网站,期间将举办专题学术报告.

报告题目：Stochastic resetting: applicationsin economics and machine learning

时间：2025年11月4日10:00

地点：理工楼631

报告摘要：Stochastic resetting to a reference state acts as a simple, tunable mechanism for stabilization and optimization in complex systems. In economics, augmenting income and wealth dynamics with resets helps control non-ergodicity, producing regimes where inequality stabilizes and mobility can be quantified more granularly via first-passagetime metrics beyond transition matrices. Theory shows that, for fixed randomness, larger driftlowers the optimal resetting rate that minimizes the mean first-passagetime (MFPT), whereas for a fixed drift, greater

randomness raises this optimum. Resetting is beneficial when diffusion is sufficiently strong relative to the drift, a fact that can be expressed through a dimensionless drift-diffusion ratio (Péclet number). Interestingly, recent research in machine learning shows that the same principle appears in neural network training with label noise: resetting model parameters to a checkpoint mitigates memorization and improves generalization, with performance typically U-shaped as a function of the reset probability. Since performance is maximized at an optimal reset probability, an adaptive schedule that adjusts this probability during training eliminates manual tuning and achieves performance comparable to or exceeding the best fixed setting.

欢迎广大师生光临！

报告人简介

Petar Jolakoski is a PhD student at the Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University in Skopje, Macedonia. He holds a MSc in Sta*s*cal Methods from the Faculty of Economics in Skopje. He has a teaching experience at the Faculty of Economics Skopje, as part of the Chair of Mathematics and Statistics in the period from 2019 to 2021, where he participated in the preparation and implementation of laboratory exercises in Mathematics, Statistics, Financial Econometrics and Statistical Quality Control. His teaching experience continues as a lecturer at a private college in Skopje from 2022 to 2025 where he gave lectures and laboratory exercises in Linear Algebra and Statistics and Probability. Since 2021 he has been working as a junior research associate at the Macedonian Academy of Sciences and Arts on projects funded by the German Science Foundation (DFG) and the Alliance of International Science Organizations (ANSO). He has published over 15 papers in peer-reviewed, impact-factor journals. His current work spans statistical physics and its applications to socio-economic systems and machine learning.

伟德体育伟德体育官方网站

萃英公司

2025年10月30日