Seminar: The Random Neural Network and G-Networks: Mathematical Tools for Machine Learning and System Performance Evaluation

When: Friday, April 11, 2025, 3:00 PM - 4:00 PM
Where: Peter Landin Building 4.24, Mile end

Abstract: Since the 1970s, mathematical models known as queueing networks have been commonly used for the performance analysis and optimisation of large-scale computer systems and networks, resulting in computationally efficient software tools from the 1980's such as QNAP (a commercial output of INRIA's MEVAL project founded by the speaker) and RESQ (IBM Yorktown Heights). In this informal presentation, we will recall the earliest and most widely used "Gordon and Newell" and "Jackson" queueing network models and then describe two substantial generalisations. These generalisations will be illustrated with recent examples of their use in the performance optimisation of distributed systems and machine learning.

Speaker bio: Erol Gelenbe FACM & FIEEE, Institute of Theoretical & Applied Informatics, Polish Academy of Sciences