Organizing Committee
Abstract

This workshop is concerned with sampling challenges, modeling and simulation for data-rich applications in high dimensions. It brings together mathematicians, statisticians and computational scientists to explore the interplay between computational applied mathematics and data science. On the agenda will be novel developments in the study of complex phenomena based on data-analytic techniques, such as efficient calculation of ergodic (long term) averages and statistical inference under a wide range of geometric, physical and analytical constraints.

In applied mathematics and computational science, in particular in molecular modeling, image analysis and geosciences, among others, many objects of interest are high-dimensional and stochastic, and a wide variety of techniques have been developed for sampling and approximating the quantities of interest. Similar issues arise in the area of data science and statistical modeling, where learning problems in the presence of high-dimensional data require efficient computational algorithms for sampling and approximation.

The workshop will focus on recent advances in the design of rigorous discrete-dynamics based sampling approaches, algorithms development for large-scale data analysis and stochastic dynamical systems, scalable and rigorous numerical methods for stochastic differential equations and sampling from high-dimensional distributions, and exploitation of low-dimensional structures in high-dimensional data and stochastic dynamical systems for model reduction and efficient Monte-Carlo schemes. The meeting will foster the interchange and deployment of the latest methodologies for sampling and approximation.

A measure induced by a stochastic differential equation whose drift induces a limit cycle, presenting a challenge to local simulation schemes. [Photo credit: Mark Girolami]

Confirmed Speakers & Participants

Workshop Schedule

Monday, July 18, 2016
TimeEventLocationMaterials
8:30 - 8:55Registration11th Floor Collaborative Space 
8:55 - 9:00Welcome - ICERM Director11th Floor Lecture Hall 
9:00 - 9:35Revisiting Monte Carlo Estimation - Mark Girolami, University of Warwick11th Floor Lecture Hall 
9:45 - 10:20Sequential matrix completion - Sergio Bacallado, Universiy of Cambridge11th Floor Lecture Hall
10:30 - 11:00Coffee/Tea Break11th Floor Collaborative Space 
11:00 - 11:35Fast Robustness Quantification with Variational Bayes - Tamara Broderick, Massachusetts Institute of Technology11th Floor Lecture Hall
11:45 - 12:20Adaptive sparse quadrature for high-dimensional integration with Gaussian distribution- application to Bayesian inverse problems. - Peng Chen, The University of Texas at Austin11th Floor Lecture Hall
12:30 - 3:00Break for Lunch / Free Time  
3:00 - 3:35Data-driven stochastic model reduction - Fei Lu, University of California, Berkeley11th Floor Lecture Hall
3:45 - 4:20Mathematics for data-driven modeling - The science of crystal balls - Yannis Kevrekides, Princeton University11th Floor Lecture Hall
4:30 - 6:00Welcome Reception11th Floor Collaborative Space 
Tuesday, July 19, 2016
TimeEventLocationMaterials
9:00 - 9:35Multi-fidelity information fusion algorithms for high dimensional systems and massive data-sets - George Karniadakis, Brown University11th Floor Lecture Hall
9:45 - 10:20Product form stationary distributions of stochastic reaction networks with non-mass action kinetics, and application to constrained averaging of multiscale systems - Simon Cotter, University of Manchester11th Floor Lecture Hall
10:30 - 11:00Coffee/Tea Break11th Floor Collaborative Space 
11:00 - 11:35Statistical inference of stochastic single-cell and single-molecule dynamics - Heinz Koeppl, Technische Universität Darmstadt11th Floor Lecture Hall
11:45 - 12:20Measure transport, variational inference, and low-dimensional maps - Youssef Marzouk, Massachusetts Institute of Technology11th Floor Lecture Hall
12:30 - 2:15Break for Lunch / Free Time  
2:15 - 2:50Stochastic integrators for multiscale and ergodic dynamical systems - Assyr Abdulle, École Polytechnique Fédérale de Lausanne11th Floor Lecture Hall
3:00 - 3:35Stratification for Markov chain Monte Carlo sampling - Jonathan Weare, University of Chicago11th Floor Lecture Hall
3:45 - 4:15Coffee/Tea Break11th Floor Collaborative Space 
4:15 - 5:00Lightning Talks11th Floor Lecture Hall 
Wednesday, July 20, 2016
TimeEventLocationMaterials
9:00 - 9:35Efficient methods for estimating stochastic gradients and sensitivity indices in complex stochastic dynamics - Markos Katsoulakis, University of Massachusetts, Amherst11th Floor Lecture Hall
9:45 - 10:20Quasi-Reliable Estimates of Effective Sample Size - Robert Skeel, Purdue University11th Floor Lecture Hall
10:20 - 10:30Group Photo in Lecture Hall11th Floor Lecture Hall 
10:30 - 11:00Coffee/Tea Break11th Floor Collaborative Space 
11:00 - 11:35Ergodic Stochastic Differential Equations and Sampling- A numerical analysis perspective - Konstantinos Zygalakis, University of Edinburgh11th Floor Lecture Hall
11:45 - 1:50Break for Lunch / Free Time  
2:00 - 2:35Ensemble preconditioning schemes for Bayesian inverse problems - Charles Matthews, University of Chicago11th Floor Lecture Hall
2:45 - 3:20Accurate sampling algorithms for constraints and noisy gradients - Benedict Leimkuhler, University of Edinburgh11th Floor Lecture Hall 
3:30 - 5:00Poster Session and Coffee/Tea Break11th Floor Collaborative Space 
Thursday, July 21, 2016
TimeEventLocationMaterials
9:00 - 9:35Variational Hamiltonian Monte Carlo - Babak Shababa, The Center for Statistical Consulting11th Floor Lecture Hall
9:45 - 10:20Scalable and Efficient MCMC Algorithms for Complex Posteriors - Yian Ma, University of Washington11th Floor Lecture Hall
10:30 - 11:00Coffee/Tea Break11th Floor Collaborative Space 
11:00 - 11:35Sub-Gaussian estimators of the mean of a random matrix with entries possessing only two moments - Stanislav Minsker, University of Southern California11th Floor Lecture Hall
11:45 - 12:20Adaptive two-stage integrators for sampling algorithms based on Hamiltonian dynamics - Elena Akhmatskaya, Basque Center for Applied Mathematics - BCAM11th Floor Lecture Hall
12:30 - 2:15Break for Lunch / Free Time  
2:15 - 2:50Scalable Bayesian Inference with Hamiltonian Monte Carlo - Michael Betancourt, University of Warwick11th Floor Lecture Hall
3:00 - 3:35Computing ergodic limits for SDEs - Michael Tretyakov, The University of Nottingham11th Floor Lecture Hall
3:45 - 4:15Coffee/Tea Break11th Floor Collaborative Space 
4:15 - 5:00Panel Discussion11th Floor Lecture Hall 
Friday, July 22, 2016
TimeEventLocationMaterials
9:00 - 9:35Geometric Methods for the Approximation of High-dimensional Dynamical Systems - Mauro Maggioni, Johns Hopkins University11th Floor Lecture Hall
9:45 - 10:20Scalable algorithms for Markov process parameter inference - Darren Wilkinson, Newcastle University11th Floor Lecture Hall
10:30 - 11:00Coffee/Tea Break11th Floor Collaborative Space 
11:00 - 11:35Covariance-controlled adaptive Langevin thermostat for large-scale Bayesian sampling - Xiaocheng Shang, Brown University11th Floor Lecture Hall
11:45 - 12:20Irreversible langevin samplers and variance reduction- a large deviations approach and diffusion on graphs - Konstantinos Spiliopoulos, Boston University11th Floor Lecture Hall
12:30 - 2:15Break for Lunch / Free Time  
2:15 - 2:50Simple ergodic variants of the Hamiltonian Monte Carlo method - Jesus Sanz-Serna, Universidada Carlos III de Madrid11th Floor Lecture Hall
3:00 - 3:45Wrapup Q&A11th Floor Lecture Hall 

Lecture Videos