ANC Workshop - Chris Williams and Ioannis Pisokas

Speaker: Chris Williams

Title: Inference for Generative Capsule Models

Abstract:

Capsule networks (see e.g. Hinton et al, 2018) aim to encode knowledge and reason about the relationship between an object and its parts.  In this paper we focus on a clean version of this problem, where data is generated from multiple geometric objects (e.g. triangles, squares) at arbitrary translations, rotations and scales, and the observed datapoints (parts) come from the corners of all objects, without any labelling of the objects.  We specify a generative model for this data, and derive a variational algorithm for inferring the transformation of each object and the assignments of points to parts of the objects.  Recent work by Kosiorek et al (2019) has used amortized inference via stacked capsule autoencoders (SCA) to tackle this problem---our results show that we significantly outperform them.  We also investigate inference for this problem using a RANSAC-type algorithm.

Joint work with Alfredo Nazabal.

Speaker: Ioannis Pisokas

Title: How Do Insects Remember Where Their Nest Is?

Abstract:

Insects are impressive navigators; their brains enable them to travel several kilometres searching for food and then return to their nest even without having access to visual landmarks or pheromone trails. Previous work has shown that they achieve this by remembering the direction and distance to their nest at all times. However, how this memory is updated and maintained remains an open question. Unlike other types of memory, it possesses a unique combination of short-term and long-term memory characteristics. I will present our latest findings about the neuronal basis of this memory and how anaesthesia affects it.