Fakulta matematiky, fyziky
a informatiky
Univerzita Komenského v Bratislave

Seminár z umelej inteligencie - Enda Howley, Patrick Mannion (6.11.2023)

v pondelok 6.11.2023 o 14:10 hod. v miestnosti I/9 aj online


02. 11. 2023 12.04 hod.
Od: Igor Farkaš

Na programe seminára z umelej inteligencie sú dve prezentácie kolegov zo School of Computer Science, University of Galway, Írsko, ktorí navštívia fakultu v rámci projektu ENLIGHT. University of Galway je jednou z partnerských univerzít aliancie. 

Prednášajúci: Assoc. Prof. Enda Howley (University of Galway, Írsko)

Názov: An overview of using ML and simulation on a range of problem areas; Data Center Energy Optimisation, Traffic Signal Control, Cancer Cells and Cybersecurity

Abstrakt: 
In this short presentation, I will briefly introduce the School of Computer Science at the University of Galway, and then give an overview of a range of problem domains where we have been successfully applying machine learning techniques to provide new insights. Primarily reinforcement learning and neural networks have been used, but we are always very open to using a range of possible tools to address challenges in these problem domains. We have explored challenges such as how virtual machines can be consolidated in a cloud data centre, as we know this can be a major drain on overall energy usage. We have explored aspects traffic signal control and the uses of reinforcement learning in that context. More recently we have been exploring problem domains in the diverse areas of cybersecurity, protein peptide identification and even cancer cell simulation.


Prednášajúci: Dr. Patrick Mannion (University of Galway, Írsko)

Názov: Multi-objective reinforcement learning: an overview of theory and applications

Abstrakt: 
This presentation will give an overview of the field of multi-objective reinforcement learning. The talk will cover the topic of reward design for AI systems and present arguments for and against multi-objective reward design. Real world examples/use cases will be used to motivate the development of multi-objective approaches, and to highlight the shortcomings of traditional single-objective reinforcement learning. Use cases covered will include public health decision making/pandemic management, 5G cellular networks and compute resource scheduling on embedded systems.

Seminár sa uskutoční v pondelok 6.11.2023 o 14:10 hod. v miestnosti I/9 aj online cez MS Teams