The following schedule is based on Japanese standard time.
Date: 16th July
Time: 15:00 - 16:00 JST
In person location: 3rd floor of Building 11, Seikei University
Speaker: Hsin-Hsiung Huang, University of Central Florida
Title: Learning Hidden Roles in Multilayer Networks: Uncertainty Quantification with Applications to U.S. Air Travel
Abstract: Network data arise in many modern applications, including transportation systems, social networks, biological networks, and communication systems. In many cases, we observe not just one network, but a collection of related networks over time or across different conditions. For example, U.S. airport traffic can be viewed as a monthly sequence of networks, where airports are nodes and flights between airports are edges.
A central question is how to identify the hidden roles of nodes in such multilayer networks. An airport, for instance, may behave partly like a regional hub and partly like a national connector. These roles may also change in strength over time, especially during major disruptions such as the COVID-19 pandemic. In this talk, I will introduce a statistical framework for learning such mixed roles while also quantifying uncertainty in the results. The method combines fast spectral ideas with a Bayesian-inspired variational approximation, making it computationally practical for large multilayer networks.
The talk will focus on the intuition behind the model, why uncertainty quantification is important, and what can be learned from real data. I will illustrate the method using a U.S. airport transportation network from 2016 to 2021, comparing the network structure before and during the COVID-19 period. I will also briefly discuss the theoretical guarantees behind the method, while keeping the main presentation accessible to students and researchers who may not have extensive background in Bayesian asymptotic theory.
Could you please let me know the expected length of the talk, including or excluding questions? Also, it would be helpful to know the likely audience composition. For example, should I prepare the talk with more emphasis on intuition, methodology, and applications for a broader audience, or should I include more theoretical details for a statistics/probability audience?
オンライン参加希望の方はフォームに登録をお願いします．If you plan to attend online, kindly register by using the link:
Virtual: https://us06web.zoom.us/meeting/register/-TVJjPdxRHSFnqNWfIbarA
Date: 13th July
Time: 15:00 - 16:00 JST
In person location: TBA
Speaker: Ning Ning (Patricia), Texas A&M University
Title: Bayesian Inference for Partially Observed McKean-Vlasov SDEs with Full Distribution Dependence
Abstract: McKean–Vlasov stochastic differential equations (MVSDEs) describe systems whose dynamics depend on both individual states and the population distribution, and they arise widely in neuroscience, finance, and epidemiology. In many applications the system is only partially observed, making inference very challenging when both drift and diffusion coefficients depend on the evolving empirical law. This paper develops a Bayesian framework for latent state inference and parameter estimation in such partially observed MVSDEs. We combine time-discretization with particle-based approximations to construct tractable likelihood estimators, and we design two particle Markov chain Monte Carlo (PMCMC) algorithms: a single-level PMCMC method and a multilevel PMCMC (MLPMCMC) method that couples particle systems across discretization levels. The multilevel construction yields correlated likelihood estimates and achieves mean square error $(O(\varepsilon^2))$ at computational cost $(O(\varepsilon^{-6}))$, improving on the $(O(\varepsilon^{-7}))$ complexity of single-level schemes. We address the fully law-dependent diffusion setting which is the most general formulation of MVSDEs, and provide theoretical guarantees under standard regularity assumptions. Numerical experiments confirm the efficiency and accuracy of the proposed methodology.
オンライン参加希望の方はフォームに登録をお願いします．If you plan to attend online, kindly register by using the link:
Virtual: https://us06web.zoom.us/meeting/register/Q7lpj_lHRM6EAGwMxwb3MA
Date: 2nd March
Time: 15:00 - 16:00 JST
In person location: Seminar Room B, Building No. 6, Faculty of Engineering, Hongo Campus, University of Tokyo
Speaker: Jeffrey S. Rosenthal, University of Toronto
Title: Optimising MCMC Tempering Algorithms
Abstract: Markov chain Monte Carlo (MCMC) algorithms are very popular in Bayesian statistics, but have difficulties moving between widely separated modes of the target distribution. A common solution is Simulated or Parallel Tempering algorithms, which use fractional powers of the target density to facilitate inter-mode transitions. This leads to questions about what choices to make re temperature spacings, balancing within- and between-temperature updates, and using reversible or non-reversible update schemes. In this talk, we present some theoretical results about how to optimise these choices to maximise the efficiency of the algorithm.
オンライン参加希望の方はフォームに登録をお願いします．If you plan to attend online, kindly register by using the link:
Virtual: https://us06web.zoom.us/meeting/register/kzv95m3USX63091LfJ-Q1A