Seminars will take place via Microsoft Teams, with a meeting link to be shared via the seminar-announce emails. Unless otherwise stated, from 15.00 to 16.00 on
Friday afternoons during semester time, followed by informal discussions.
| Date | Speaker | Title/Abstract |
|---|---|---|
| Friday 17 September |
||
| Friday 24 September |
Wen-shin Lee | Sparse Interpolation: design sparse antenna arrays for estimating directions of arriving signals Estimating the directions of simultaneously arriving signals plays an important role in radar, remote sensing, radio frequency interference mitigation, wireless networks, machine perception of unmanned aerial vehicles or self-driving cars. In signal processing, antenna array systems have been designed to solve the problem of estimating the direction of arrival (DOA). A main constraint in designing regularly spaced antenna systems is the spatial Nyquist criterion, which requires the space between two sensors to be less than half of the signal wavelength. A disadvantage of densely spaced antenna elements is the effect of mutual coupling, normally reduced through costly extensive calibration of the system. Using a regularly spaced antenna system for DOA estimation can be formulated as an exponential analysis problem, which can be tackled by the classical Prony method from approximation theory. Interestingly, the Ben-Or/Tiwari sparse interpolation algorithm in computer algebra is closely related to Prony's method. This connection has led to a major development in exponential analysis that can circumvent the Nyquist constraint, hence allow us to completely remove the dense Nyquist spacing requirement for DOA in antenna design. This is joint work with Annie Cuyt, Ferre Knaepkens, Dirk I. L. de Villiers. |
| Friday 1 October |
||
| Friday 8 October |
||
| Friday 15 October |
Vincenzo Crescimanna | An information-theoretic introduction to representation learning Representation Learning (RL)- the process of learning useful descriptions of the visible data - is a machine learning field with the most interest in recent years. Indeed, via RL it is possible to understand and extract the most relevant information from a large dataset or understand how to improve deep neural networks. The goal of the talk is to give a general introduction to RL, trying to explain its relevance, the challenge in defining an optimal representation, and the way to learn these representations. In particular, following an information-theoretic approach, we will describe a possible optimal definition and a learning principle to get such a representation. |
| Friday 22 October |
||
| Friday 29 October |
No seminar | UG Reading week |
| Friday 5 November |
No seminar | PG Reading week |
| Friday 12 November |
||
| Friday 19 November |
Sarah Thomson | Predicting and Improving Vocational Rehabilitation Outcomes Vocational rehabilitation aims to assist people in re-joining the workforce and maintaining employment. A large amount of data can be collected during this process, and the ADAPT (Automated Dynamic Adaptive Personalised Treatments) consortium seeks to harness that data to help predict and improve outcomes for future patients. In this work, we have already used machine learning models to predict patient outcomes with success. Our next stage focuses on improving patient outcomes, and we have been prototyping a system for that over the past year. The system is called the Pathway Generator (TPG). TPG aims to personalise and optimise treatment pathways through rehabilitation by learning from historical data and using an evolutionary algorithm to optimise. In this way, we hope that TPG can serve as a decision-assisting tool for professionals planning treatments. This talk will introduce the domain, elucidate the prototype optimiser system, and present some preliminary results. |
| Friday 26 November |
No seminar | Internal event |
| Friday 3 December |
||
| Friday 10 December |
Computing Science and Mathematics
Faculty of Natural Sciences
University of Stirling
Stirling
FK9 4LA
Scotland
UK
+44 01786 467421
+44 01786 464551
secretaries@cs.stir.ac.uk
Twitter: @csmstir
Facebook: StirUniCSM