| Borislav Ikonomov and Chris Williams |
| Tue 30 Jan 2018, 11:00 - 12:00 |
| IF 4.31/4.33 |
If you have a question about this talk, please contact: Gareth Beedham (gbeedham)
Borislav Ikonomov
Title: Addressing computational challenges in computer vision and Bayesian perception using approximate Bayesian computation
Abstract: I will discuss my current work on using approximate Bayesian computation (ABC) - a likelihood-free inference method - to do parameter estimation of visual scenes. The main advantage of this approach would be the ability to produce a full and general probability distribution over the parameters rather than a point estimate, which would allow the model to detect multi-modalities (e.g. ambiguities) in the images. I will then connect this to the cognitive model of Bayesian perception and briefly mention why likelihood-free inference could be an interesting tool to use in cognitive science as well as in computer vision.
Chris Williams
Title: Model Criticism in Latent Space
Abstract:
Model criticism is usually carried out by assessing if replicated data generated under the fitted model looks similar to the observed data, see e.g. Gelman, Carlin, Stern, and Rubin (2004, p. 165). This paper presents a method for latent variable models by pulling back the data into the space of latent variables, and carrying out model criticism in that space. Making use of a model's structure enables a more direct assessment of the assumptions made in the prior and likelihood. We demonstrate the method with examples of model criticism in latent space applied to ANOVA, factor analysis, linear dynamical systems and Gaussian processes.
Joint work with Sohan Seth and Iain Murray.