Algorithms & Computationally Intensive Inference seminars
The seminars will take place on Fridays 1 pm UK time in room MB0.08 in a hybrid format.
2021-2022 Organisers: Alice CorbellaLink opens in a new window and Lyudmila GrigoryevaLink opens in a new window
If you would like to speak, or you want to be included in any emails, please contact one of the organisers.
Website URL: www.warwick.ac.uk/compstat
Mailing List Sign-Up: http://mailman1.csv.warwick.ac.uk/mailman/listinfo/algorithmseminar
Mailing List: algorithmseminar@listserv.csv.warwick.ac.uk (NB - only approved members can post)
2021/22 Term 3
The list of firmly confirmed speakers.
Date | Speaker | Title | F2F | Slides | Video |
week 1 29/04 | No seminar | ||||
week 2 06/05 | Sinho ChewiLink opens in a new window | Improved dimension dependence for MALA and lower bounds for sampling | |||
Abstract: The optimal scaling literature predicts that the mixing time of the Metropolis-adjusted Langevin Algorithm (MALA) scales as d^{1/3}, where d is the dimension. However, the scaling limit requires stringent assumptions and is asymptotic in nature. In this work, we improve the state-of-the-art non-asymptotic mixing time bound for MALA on the class of log-smooth and strongly log-concave distributions from O(d) to O(d^{1/2}), under the additional assumption of a warm start; moreover, our bound is sharp. Our proof introduces a new technique based on a projection characterization of the Metropolis adjustment which reduces the study of MALA to the discretization analysis of the Langevin SDE. Afterwards, I will briefly discuss the elusive problem of obtaining lower bounds for sampling, including recent work which establishes the first tight complexity bound for sampling in one dimension. This is joint work with Kwangjun Ahn, Xiang Cheng, Patrik Gerber, Thibaut Le Gouic, Chen Lu, and Philippe Rigollet. | |||||
week 3 13/05 | Simon SpencerLink opens in a new window | Accelerating adaptation in MCMC algorithms | |||
Abstract: The Metropolis-Hastings random walk algorithm remains popular with practitioners due to the wide variety of situations in which it can be successfully applied and the extreme ease with which it can be implemented. Adaptive versions of the algorithm use information from the early iterations of the Markov chain to improve the efficiency of the proposal. In this talk I will describe how to reduce the number of iterations needed to adapt the proposal to the target, which is particularly important when the likelihood is time-consuming to evaluate. First, the accelerated shaping algorithm is a generalisation of both the adaptive proposal and adaptive Metropolis algorithms. It is designed to remove misleading information from the estimate of the covariance matrix of the target. Second, the accelerated scaling algorithm rapidly changes the scale of the proposal to achieve a target acceptance rate. Finally, I will show how the same ideas can be applied to efficiently update parameter vectors with Dirichlet priors. | |||||
week 4 20/05 | Sida ChenLink opens in a new window | Bayesian spline-based hidden Markov models with applications to activity acceleration data and sleep analysis | |||
Abstract: B-spline-based hidden Markov models use B-splines to specify the emission distributions and offer a more flexible modelling approach to data than conventional parametric HMMs. We introduce a Bayesian framework for inference in these models where the number of states may be unknown along with other model parameters. A trans-dimensional Markov chain inference algorithm is proposed to identify a parsimonious knot configuration of the B-splines while model selection regarding the number of states can be performed based on the marginal likelihood within a parallel sampling framework. Using an extensive simulation study, we establish the superiority of our proposed methodology in comparison to alternative approaches. We will also present an extension of the basic model - a novel hierarchical conditional HMM to analyse human accelerator activity data for circadian and sleep modelling.
|
|||||
week 5 27/05 | Lucia PaciLink opens in a new window | ||||
week 6 3/06 | BANK HOLIDAY | ||||
week 7 10/06 | Gareth Roberts | Biased Draws and Corrections | |||
Abstract: Draws for major sporting events are often televised and carried out in a sequential fashion to maximise excitement and to increase anticipation for the sporting event itself. In this regard, organisations such as FIFA and UEFA in football have been highly successful. However these draw procedures are also often subject to constraints which make the problem of simulating a fair draw (ie uniform over all feasible draws which satisfy the constraints) difficult to achieve using a sequential procedure. For example the recent FIFA World Cup draw imposed geographical constraints to avoid countries from the same continental confederation (apart from Europe) playing each other in the group stages. Recent draws by FIFA and UEFA have all (to a greater or lesser extent) been biased. From a computational statistical perspective, they have been based on approximate SMC procedures. This talk will investigate these biases and suggest practical solutions which respect the desire to unveil such a draw in a sequential fashion. The main focus will be on the FIFA 2022 World Cup draw. This is joint work with Jeffrey RosenthalLink opens in a new window | |||||
week 8 17/06 | Badr-Eddine Chérief-AbdellatifLink opens in a new window | ||||
week 9 24/06 | Mauro Camara-EscuderoLink opens in a new window | ||||
week 10 1/7 | End of year BBQ |
2021/22 Term 2
The list of confirmed speakers. For abstracts please see the Programme PageLink opens in a new window.
Date | Speaker | Title | F2F | Slides | Video | |
Week 1 14/01 |
Ryan MartinLink opens in a new window | Data-driven Calibration of Generalized Posterior Distributions | ||||
Week 2 21/01 |
No seminar | |||||
Week 3 28/01 |
Filippo Pagani | Numerical Zig-Zag and Perturbation Bounds on Numerical Error | ||||
Week 4 04/02 |
Emilia Pompe | Robust Inference using Posterior Bootstrap | ||||
Week 5 11/02 |
Benedict Leimkuhler | Partitioned Integrators and Multirate Training of Deep Neural Networks | ||||
Week 6 18/02 |
Laura Guzman RinconLink opens in a new window |
Bayesian estimation of the instant growth rate of SARS-CoV-2 positive cases in England, using Gaussian processes.
|
||||
Week 7 25/02 |
Martyn Plummer | Developing JAGS | ||||
Week 8 04/03 |
Kamran PentlandLink opens in a new window | GParareal: A time-parallel ODE solver using Gaussian process emulation | ||||
Week 9 11/03 |
No Seminar | |||||
Week 10 18/03 |
Ioannis KosmidisLink opens in a new window |
2021/22 Term 1
The list of confirmed speakers. For abstracts please see the Programme PageLink opens in a new window.
Date | Speaker | Title | F2F | Slides | Video | |
Week 1 08/10 | Lorenzo PacchiardiLink opens in a new window | |||||
Week 2 15/10 | Internal jam session: 2-5 minutes brief intro to the ongoing research projects or some new ideas | |||||
Week 3 22/10 | Petros DellaportasLink opens in a new window | Negligible-cost Variance Reduction for Metropolis-Hastings Chains | ||||
Week 4 29/10 | Sanmitra GhoshLink opens in a new window | Variational Inference for Nonlinear ODEs | ||||
Week 5 05/11 | Azadeh KhaleghiLink opens in a new window | On Some Probabilities and Limitations of Restless Multi Armed Bandits | ||||
Week 6 12/11 | Sam PowerLink opens in a new window | Accelerated Sampling on Discrete Spaces with Non-Reversible Markov Jump Processes | ||||
Week 7 19/11 | Marta CatalanoLink opens in a new window | A Wasserstein Index of Dependence for Bayesian Nonparametric modeling | ||||
Week 8 26/11 | Edward IonidesLink opens in a new window | Bagging and Blocking: Inference via Particle Filters for Interacting Dynamic Systems | ||||
Week 9 03/12 | Accurate and Efficient Numerical Methods for Molecular Dynamics and Data Science Using Adaptive Thermostats | |||||
Week 10 10/12 | Lionel Riou-DurandLink opens in a new window |
Metropolis Adjusted Underdamped Langevin Trajectories (jointly with Jure VogrincLink opens in a new window (University of Warwick)) |
2021/22 Term 2
The list of confirmed speakers. For abstracts please see the Programme PageLink opens in a new window.
Date | Speaker | Title | F2F | Slides | Video | |
Week 1 14/01 |
Ryan MartinLink opens in a new window | Data-driven Calibration of Generalized Posterior Distributions | ||||
Week 2 21/01 |
No seminar | |||||
Week 3 28/01 |
Filippo Pagani | Numerical Zig-Zag and Perturbation Bounds on Numerical Error | ||||
Week 4 04/02 |
Emilia Pompe | Robust Inference using Posterior Bootstrap | ||||
Week 5 11/02 |
Benedict Leimkuhler | Partitioned Integrators and Multirate Training of Deep Neural Networks | ||||
Week 6 18/02 |
Laura Guzman RinconLink opens in a new window |
Bayesian estimation of the instant growth rate of SARS-CoV-2 positive cases in England, using Gaussian processes.
|
||||
Week 7 25/02 |
Martyn Plummer | Developing JAGS | ||||
Week 8 04/03 |
Kamran PentlandLink opens in a new window | GParareal: A time-parallel ODE solver using Gaussian process emulation | ||||
Week 9 11/03 |
No Seminar | |||||
Week 10 18/03 |
Ioannis KosmidisLink opens in a new window |