Skip to main content

Abstracts


Talks

  • Louis Aslett (Durham) - "Cryptography and Statistics: a short introduction" short course
    • Abstract: The security of our data is an ever present and growing concern. Since at least 1500BC humans have turned to methods of cryptography to protect sensitive information, but the modern demands of statistical modelling require more than simply storage or transport of sensitive information: we must also be able to compute on the underlying data. The first part of this short course will carefully introduce cryptography in a manner accessible to statisticians and explain how recent homomorphic encryption schemes can provide the kind of security required to facilitate model building with strong security guarantees, as well as how shared computation can improve this to provide information theoretic levels of security. This will be followed by a brief overview of how these methods have been deployed to solve statistical problems in the literature, with a deeper treatment of recent research that develops new statistical methodology to enable model building within these constrained cryptographic environments.
  • Owen Jones (Cardiff) - "Stochastic Simulation and Optimisation" short course
    • Talk 1 / 3: "Stochastic Gradient methods"
      • Abstract: We start with a brief taxonomy of stochastic optimisation methods, before having a closer look at Stochastic Gradient methods. In particular we look at the classic result of Kiefer and Wolfowitz and the Simultaneous Perturbation method of Spall. Finally we look at some refinements available when we apply these to simulated data.
    • Talk 2 / 3: "Cross-entropy"  
      • Abstract: The Cross-Entropy (CE) method started as an adaptive importance sampling scheme for estimating rare event probabilities, before being successfully applied to a variety of combinatorial optimisation problems. It is a model based stochastic search technique and requires a parameterised sampling distribution. We look at some examples and discuss some convergence results.
    • Talk 3 / 3: "Simulation based estimation"
      • Abstract: Often we are faced with estimating a model with an intractable likelihood, but which we can none-the-less simulate. We look at how we can use simulations to carry out the equivalent of method of moments, maximum likelihood, or Bayesian estimation.
  • Mark Briers (ATI) - "Statistical Challenges in Cyber Security"
    • Abstract: With the realisation that Cyber attack presents a significant risk to an organisation's reputation, efficiency, and profitability, there has been an increase in the instrumentation of networks; from collecting netflow data at routers, to host-based agents collecting detailed process information. To spot the potential threats within a Cyber environment, a large community of researchers have produced many exciting innovations, aligned with such data. Much of this research has been focused around "data driven" techniques, and does not often fuse data from multiple sources. Moreover, incorporation of threat actors' behaviours and motivations (as specified by Cyber security experts) is often non-existent. In this talk, I will present an overview of the statistical challenges facing the Cyber domain, and demonstrate the use of two-filter smoothing within a state-space modelling context for the characterisation of user behaviour within a point-process model.
  • Christian Robert (Dauphine) - "ABC convergence and misspecification"
    • Abstract: Approximate Bayesian computation is becoming an accepted tool for statistical analysis in models with intractable likelihoods. With the initial focus being primarily on the practical import of this algorithm, exploration of its formal statistical properties has begun to attract more attention. In this paper we consider the asymptotic behaviour of the posterior distribution obtained by this method. We give general results on: (i) the rate at which the posterior concentrates on sets containing the true parameter (vector); (ii) the limiting shape of the posterior; and (iii) the asymptotic distribution of the ensuing posterior mean. These results hold under given rates for the tolerance used within the method, mild regularity conditions on the summary statistics, and a condition linked to identification of the true parameters. The issue of model mis-specification in ABC settings will also be considered. [This is joint work with D. Frazier, G. Martin, & J. Rousseau]
  • Patrick Rubin-Delanchy (Bristol) - "Mathematical progress on the connection between spectral embedding and network models used by the probability, statistics and machine-learning communities"
    • Abstract: In this talk, I give theoretical and methodological results, based on work spanning Johns Hopkins, the Heilbronn Institute for Mathematical Research, Imperial and Bristol, regarding the connection between various graph spectral methods and commonly used network models which are popular in the probability, statistics and machine-learning communities. An attractive feature of the results is that they lead to very simple take-home messages for network data analysis: a) when using spectral embedding, consider eigenvectors from both ends of the spectrum; b) when implementing spectral clustering, use Gaussian mixture models, not k-means; c) when interpreting spectral embedding, think of "mixtures of behaviour" rather than "distance". Results are illustrated with cyber-security applications.
  • Sara Wade (Warwick) - "Pseudo-marginal Bayesian inference for supervised Gaussian process latent variable models"
    • Abstract: We introduce a Bayesian framework for inference with a supervised version of the Gaussian process latent variable model. The framework overcomes the high correlations between latent variables and hyperparameters by using an unbiased pseudo estimate for the marginal likelihood that approximately integrates over the latent variables. This is used to construct a Markov Chain to explore the posterior of the hyperparameters. We demonstrate the procedure on simulated and real examples, showing its ability to capture uncertainty and multimodality of the hyperparameters and improved uncertainty quantification in predictions when compared with variational inference.

Posters

  • Francois-Xavier Briol (Warwick) - "Bayesian Quadrature for Multiple Related Integrals"
  • Mathias Cronjager (Oxford) - "How many ways can we explain the same set of genetic sequences?"
  • David Hartley (Warwick) - "A Bayesian approach to structured expert judgement"
  • Jack Jewson (Warwick) - "Robust Bayesian Updating
  • Kim Kenobi (Aberystwyth) - "The dolphins of Cardigan Bay - modelling dolphin sightings 2005-15"
  • Giulio Morina (Warwick) - "From a Bernoulli Factory to a Coin Enterprise via Perfect Sampling of Markov Chains"
  • Francesco Sanna Passino (Imperial) - "Bayesian methods for separating human and automated activity in large computer networks"
  • David Selby (Warwick) - "Ranking influential communities in networks"
  • Alex Terenin (Imperial) - "Polya Urn Latent Dirichlet Allocation: using sparsity to accelerate MCMC in natural language processing"
  • George Vasdekis (Warwick) - "The Zig-Zag Process"
  • Rachel Wilkerson (Warwick) - "A Suite of Bayesian Diagnostics for the Chain Event Graph"