Models for Reinforcement Learning and Design of a Soft Robot Inspired by Drosophila Larva

Abstract Designs for robots are often inspired by animals. Robots can mimic animals’ mechanics, motions, behaviours and learning. These properties and functions have been well studied in fruit fly, Drosophila. This thesis explores the Drosophila larva their neural circuits for operant learning and motor system for locomotion. The former is for bio-inspired robot learning and the latter is used for soft robot system design. A synaptic plasticity model and a neural circuit model for operant learning are proposed based on research of Drosophila larva.

Training Spiking Neural Networks with Surrogate Gradients

Abstract Computation in the brain is largely performed by spiking neural networks. However, currently, we neither understand how spiking networks in neurobiology compute nor how to instantiate such capabilities in artificial network models. In my talk, I will revisit the problem of supervised learning in multi-layer and recurrent spiking neural networks. Crucially, standard gradient-based optimization methods fail in the spiking setting because gradients cannot propagate through the threshold nonlinearity of most spiking neuron models.

Probabilistic Deep Learning: Foundations, Applications and Open Problems

Abstract Advances in deep generative models are at the forefront of deep learning research because of the promise they offer for allowing data-efficient learning, and for model-based reinforcement learning. In this talk I’ll review the foundations of probabilistic reasoning and generative modeling. I will then introduce modern approximations which allow for efficient large-scale training of a wide variety of generative models, demonstrate a few applications of these models to density estimation, missing data imputation, data compression and planning.

Community Structure Validation in Networks

Abstract High throughput technologies have led to an increased availability of data and to the need for novel statistical tools. Biological networks provide a mathematical representation of patterns of interaction between appropriate biological elements. We propose a novel approach to compare community structures in different networks. During this seminar we will try to address some open questions: How can we compare two (or more) networks and their community structures? Can we use Network Enrichment Analysis tools to do this?

Variational Inference for Stochastic Differential Equations

Abstract A stochastic differential equation (SDE) defines a random function of time, known as a diffusion process, by describing its instantaneous behaviour. As such, SDEs are powerful modelling tools in fields such as econometrics, biology, physics and epidemiology. This talk considers the common problem where SDEs involve unknown parameters which we wish to infer from partial noisy observations of the diffusion process. Parameter inference for SDEs is challenging due to the latent diffusion process.

Digital Humanities Perspectives on AI/Machine Learning

Abstract The talk will introduce some of Tobias’ work in artificial intelligence and big data. His current work firstly looks at the opportunities of AI technology to increase both the scope and complexity of researching human culture based on historical and born-digital sources. But his work also focuses on the epistemological implications of AI, including an understanding of its bias and how meaning develops through the eyes of a computer.

Are Supervised Learning Algorithms the Key to a Paradigm Shift in the Way We Measure Air Pollution?

Abstract Low cost chemical sensors may prove to be a disruptive technology for air pollution measurements. The potential for these technologies is huge, enabling measurements on previously unachievable spatial scales and providing affordable tools to help tackle one of the largest environmental health risks in the developing world. Recent academic scrutiny has highlighted several issues with the relatively simple analytical methods used in these sensors, compared with traditional monitoring equipment, and methods to overcome these challenges need to be developed before they can reach their full potential.

Scalable Unsupervised Phenotyping using Tensor Factorization

Abstract Originally purposed to streamline documentation of care, Electronic Health Records (EHRs) provide a massive amount of diverse and readily available data that can be used to tackle important healthcare problems. Clinical phenotyping is one of them, which refers to identifying patient subgroups sharing common clinically meaningful characteristics. However, there are significant challenges in using EHR data to computationally tackle this problem, related to algorithmic scalability, model interpretability and the longitudinal nature of patient data.

Advances in GANs based on the MMD

Abstract Generative adversarial networks have led to huge improvements in sample quality for image generation. But their success is hindered by both practical and theoretical problems, leading to the proposal of a huge number of alternative methods over the last few years. We study one of these alternatives, the MMD GAN, which uses a similar architecture to an original GAN but does some of its optimization in closed form, in a Hilbert space.

Bayesian Quadrature for Multiple Related Integrals

Abstract Bayesian probabilistic numerical methods are a set of tools providing posterior distributions on the output of numerical methods. The use of these methods is usually motivated by the fact that they can represent our uncertainty due to incomplete/finite information about the continuous mathematical problem being approximated. In this talk, we demonstrate that this paradigm can provide additional advantages, such as the possibility of transferring information between several numerical methods.