Ashok Maitra Memorial Lectures 2024-25

Prof. Pietro Caputo

Dipartimento di Matematica e Fisica, Università Roma Tre Largo San Murialdo 1, 00146 Roma - Italy

PROFILE
Speaker Bio

At Indian Statistical Institute, Bangalore Centre


TALK 1
Title: Functional Inequalities in Random Permutation Dynamics: Spectral Gaps and Entropy Factorizations
Date: January 03, 2025 (FRIDAY) Time: 2-3 PM
Venue: Auditorium (II floor)
Slides
Abstract: We explore functional inequalities associated with random permutation dynamics, such as random transpositions and other card shuffling algorithms. Beginning with the spectral gap of the interchange process on arbitrary weighted graphs and its generalization to hypergraphs, we then address their entropic counterparts. These entropic inequalities lead naturally to entropy factorization estimates, which have significant applications, including resolving a longstanding conjecture on the upper bound of the permanent of arbitrary matrices with nonnegative entries. Along the way, we highlight recent advancements and outline some open questions.

At Indian Statistical Institute, Bangalore Centre


TALK 2
Title: Nonlinear dynamics for spin systems: some convergence results
Date: January 03, 2025 (FRIDAY) Time: 3.15-4.15 PM
Venue: Auditorium (II floor)
Slides
Abstract: We introduce a natural class of nonlinear dynamics for combinatorial structures and spin systems, such as the Ising model. This class is based on the framework of mass action kinetics, which models the evolution of particle systems under pairwise interactions, capturing several important nonlinear models from various fields, including Boltzmann's equation, recombination in population genetics, and genetic algorithms. In the context of spin systems, this approach provides a nonlinear Monte Carlo dynamics, which appears to be much harder to analyze than its linear Markov chain counterpart. In this talk, we present a general theorem on convergence to stationarity as well as some sharp quantitative results in the high-temperature regime. The proof combines the analysis of fragmentation and branching processes with a novel coupling of high-temperature Ising models and sub-critical Erdős-Rényi random graphs. We conclude with a discussion of open problems and directions for future research. 

At Indian Statistical Institute, Kolkata


PUBLIC LECTURE
Title: Navigating Networks by Random Walks
Date: January 08, 2025 (WEDNESDAY) Time: 4.30 PM
Venue: L-Infinity Seminar Room, A. N. Kolmogorov Bhavan
Zoom Link: Join Zoom Meeting
https://zoom.us/j/97164296741?pwd=8ntYwKfmaDelCc34UikDxEocau0eRP.1
Meeting ID: 971 6429 6741
Passcode: 004232
Youtube Link: https://youtu.be/vvB9p7v_M_Q
Poster
Abstract: In the analysis of a large, complex network, exploration via random walks is often the most effective strategy available. How long does it take to traverse the entire network ? Does the walk achieve a stable equilibrium, and if so, how long is the journey to reach it ? And crucially, what does this steady state look like ? The answers to these questions are key to understanding a network's structural features and provide valuable insights for ranking systems and search algorithms. Developing a comprehensive mathematical treatment of these issues for real-world networks poses a formidable task, and a rigorous analysis has only recently started in the context of relatively simple models. In this lecture, we illustrate some promising preliminary progress in the setting of directed networks obtained from simple random graph models. The discussion includes scenarios where the walker experiences regeneration events such as teleportation, as in PageRank algorithms, or resampling of the underlying graph, as in a dynamically evolving network.

At Indian Statistical Institute, Kolkata


COLLOQUIUM
Title: Functional Inequalities in Random Permutation Dynamics: Spectral Gaps and Entropy Factorizations
Date: January 09, 2025 (THURSDAY) Time: 4.30 PM
Venue: L-Infinity Seminar Room, A. N. Kolmogorov Bhavan
Zoom Link: Join Zoom Meeting
https://zoom.us/j/93450544302?pwd=HgqQHb0uMpamUn2aRk3EZ67DKrC0rH.1
Meeting ID: 934 5054 4302
Passcode: 227718
Youtube Link: https://youtu.be/Ysb8Xocvdl8
Poster
Abstract: We explore functional inequalities associated with random permutation dynamics, such as random transpositions and other card shuffling algorithms. Beginning with the spectral gap of the interchange process on arbitrary weighted graphs and its generalization to hypergraphs, we then address their entropic counterparts. These entropic inequalities lead naturally to entropy factorization estimates, which have significant applications, including resolving a longstanding conjecture on the upper bound of the permanent of arbitrary matrices with nonnegative entries. Along the way, we highlight recent advancements and outline some open questions.

At Indian Statistical Institute, Delhi Centre


TALK 1
Title: Functional Inequalities in Random Permutation Dynamics: Spectral Gaps and Entropy Factorizations
Date: January 14th, 2025 (TUESDAY) Time: 3.30-4.30 PM
Venue: Auditorium
Abstract: We explore functional inequalities associated with random permutation dynamics, such as random transpositions and other card shuffling algorithms. Beginning with the spectral gap of the interchange process on arbitrary weighted graphs and its generalization to hypergraphs, we then address their entropic counterparts. These entropic inequalities lead naturally to entropy factorization estimates, which have significant applications, including resolving a longstanding conjecture on the upper bound of the permanent of arbitrary matrices with nonnegative entries. Along the way, we highlight recent advancements and outline some open questions.

At Indian Statistical Institute, Delhi Centre


TALK 2
Title: Nonlinear dynamics for spin systems: some convergence results
Date: January 15th, 2025 (WEDNESDAY) Time: 3.30-4.30 PM
Venue: Auditorium
Abstract: We introduce a natural class of nonlinear dynamics for combinatorial structures and spin systems, such as the Ising model. This class is based on the framework of mass action kinetics, which models the evolution of particle systems under pairwise interactions, capturing several important nonlinear models from various fields, including Boltzmann's equation, recombination in population genetics, and genetic algorithms. In the context of spin systems, this approach provides a nonlinear Monte Carlo dynamics, which appears to be much harder to analyze than its linear Markov chain counterpart. In this talk, we present a general theorem on convergence to stationarity as well as some sharp quantitative results in the high-temperature regime. The proof combines the analysis of fragmentation and branching processes with a novel coupling of high-temperature Ising models and sub-critical Erdős-Rényi random graphs. We conclude with a discussion of open problems and directions for future research.