Reconciling discrete and continuous modeling for the analysis of large-scale Markov chains
Sala riunioni B, edificio ZETA - Campus Scientifico via Torino e online
Mirco Tribastone, IMT School for Advanced Studies Lucca
Markov chains are a fundamental tool for stochastic modeling across a wide range of disciplines. Unfortunately, their exact analysis is often hindered in practice due to the massive size of the state space---an infamous problem plaguing many models based on a discrete state representation. When the system under study can be conveniently described as a population process, approximations based on mean-field theory have proved remarkably effective. However, since such approximations essentially disregard the effect of noise, they may potentially lead to inaccurate estimations under conditions such as bursty behavior, separation of populations into low- and high-abundance classes, and multi-stability.
This talk will present a new analytical method, finite state expansion, that combines an accurate discrete representation of a subset of the state space with mean-field equations to improve accuracy at a user-tunable computational cost. Challenging examples drawn from queuing theory and systems biology will show how finite state expansion significantly outperforms state-of-the-art approximation techniques.
This is joint work with Luca Bortolussi, Francesca Randone, Andrea Vandin, and Tabea Waizmann.
Join Zoom Meeting
Meeting ID: 824 4886 5669
Evento finanziato dal Progetto di Ricerca di Rilevante Interesse Nazionale - PRIN 2020 “Nirvana - Noninterference and Reversibility Analysis in Private Blockchains” - N. 20202FCJMH.
Mirco Tribastone is Full Professor of Computer Science at IMT School for Advanced Studies Lucca since 2020, where he leads the research group SySMA - Systems Security Modeling and Analysis and currently serves as Deputy Rector and scientific coordinator of the Ph.D. doctoral track on Software Quality. Previously, Mirco has been Associate Professor of Computer Science at IMT since 2015, after holding a position as Associate Professor at the University of Southampton (2013-2015) and as Juniorprofessur at the Ludwig-Maximilians University of Munich (2010-2013). He received a Ph.D. in Computer Science from the University of Edinburgh in 2010. Mirco’s research interests are related to the development and analysis of models for dynamical systems, with a focus on the study of performance for computer systems and applications to complex networks, for which he has authored over 120 scientific publications. After having been principal investigator for research projects funded by the European community and the German research foundation DFG, and having served as guest professor through a DFG-funded Mercator Fellowship at the TU Braunschweig (Germany), Mirco is currently the national coordinator of a three-year project funded by the Italian Ministry of University and Research on topics concerning cyber-physical systems. Since 2020, he has been involved in technology transfer through the start-up Valuematic, of which he is founding partner.
Personal page: http://cse.lab.imtlucca.it/~mirco.tribastone/
Google Scholar profile: https://scholar.google.it/citations?user=OKbh37IAAAAJ&hl=it
L'evento si terrà in inglese
Dipartimento di Scienze Ambientali, Informatica e Statistica, Prof. Sabina Rossi