Events are All You Need
| Dates: | 30 April 2025 |
|---|
| Times: | 13:00 - 14:00 |
|---|
| What is it: | Seminar |
|---|
| Organiser: | Department of Computer Science |
|---|
| Who is it for: | University staff, Current University students |
|---|
| Speaker: | Dr. Luca Peres |
|---|
|
Abstract
The number of deployed artificial intelligence models is growing rapidly. Large language models are both compute and memory intensive in training and inference, regularly having billions of parameters and requiring petabytes of training data. In this talk, I will explore the use of events, sparse and compressed digital low-level representations able to capture physical phenomena with high fidelity, as an alternative information encoding mechanism. I will then show how event-based sensors operate and how these fit within the neuromorphic paradigm, which aims to improve scaling in machine learning, by bypassing the memory bandwidth limitations inherent in traditional hardware.
Speaker
Dr. Luca Peres
Role: Lecturer in Low Energy Comp Novel Tech
Organisation: University of Manchester
Biography: Luca obtained his PhD in computer science in 2022 at the University of Manchester exploring real-time simulations of large scale biologically-representative spiking neural networks on neuromorphic hardware. He then continued for a postdoctoral position in the Department of Computer Science, working on event-driven sensing and processing for computer vision in the scope of edge applications. Luca’s main interests are in low-power sensing and processing, with focus on event-based sensor fusion applications and in-sensor and near-sensor computing. In addition, Luca is interested in exploring novel biologically-inspired architectures with the intent of producing more reliable and low-power systems aiming to overcome the von Neumann bottleneck.
Travel and Contact Information
Find event
Kilburn_TH 1.3
Kilburn Building
Manchester
