Categories
colloquium event

Colloquium: Bio-inspired Neural Architecture Search

20 June 2022, Monday. 3PM. London Time. Online over Zoom.

Title Bio-inspired Neural Architecture Search

Speaker Wei Pang, Heriot-Watt University

Abstract In this talk I will present the work of our lab over the last three years on bio-inspired Neural Architecture Search (NAS). I will first present DeepSwarm, an NAS system that makes use of swarm intelligence. DeepSwarm has won the best paper award in UKCI 2019, and it has gained 280+ stars and 36 forks at GitHub. I will then present how we used immune-inspired approaches to optimise deep neural networks and how we did this differently with existing approaches. Finally, I will cover another evolutionary computing-based NAS specifically for optimising Graph Neural Networks for molecular property prediction as part of the EPSRC project Manufacturing Immortality.

Bio Dr Wei Pang is Associate Professor at School of Mathematical and Computer Sciences, Heriot-Watt University. He is also an academic member of the National Robotarium and Edinburgh Centre for Robotics. Previously he was Senior Lecturer until 2019 and Honorary Senior Lecturer until 2021 at Aberdeen University.He has extensive research experience in bio-inspired computing, machine learning and its applications. In particular, he is interested in artificial immune systems and swarm intelligence as applied to various machine learning problems. He has authored 130+ papers and won one best paper award and one best paper runner-up award.His research has been funded by EPSRC, Royal Society, ESRC, and Cancer Research UK. He contributed to securing and co-managing research funding in total of over £7.9M, of which £3.5M was awarded to his institutions. He is currently an investigator on three EPSRC projects: RAInS (with Aberdeen, Oxford, Cambridge), PRIME (with Cranfield, York, Glasgow, and Open), and DCEE (with Imperial and Loughborough).

Zoom link https://mdx-ac-uk.zoom.us/j/6684138396?pwd=d0I5V2JlTHVKbjlKWXZ2MW1RZ0ozQT09

Meeting ID: 668 413 8396, Passcode: mdx