Etor Arza will defend his thesis on Thursday 30 November

• The defense will take place at the Informatics Engineering Faculty in Donostia (Basque Country)

Etor Arza is a PhD student in Machine Learning at the Basque Center for Applied Mathematics (BCAM) in Bilbao, Spain. He graduated in Mathematics at the University of the Basque Country (UPV/EHU). His research interests include optimization, statistics, and benchmarking.

His thesis is entitled "Understanding Non-Convex Optimization Problems and Stochastic Optimization Algorithms" and will be supervised by Aritz Pérez (BCAM) and Ekhiñe Irurozki (Telecom Paris).

The defense will take place on Thursday 30 November at 11:00h at the Faculty of Computer Engineering of the UPV/EHU in Donostia. The defense will be available and can be followed online through the Zoom platform.

On behalf of all BCAM members, we would like to wish Etor the best of luck in his thesis defense.

Abstract 

This thesis focuses on advancing stochastic iterative heuristics optimization algorithms, aiming to compare and improve algorithms across various dimensions. Addressing the challenges of fair comparison, the research introduces methodologies to ensure a fair comparison on different hardware. In addition, it explores the limitations of traditional statistical measures and introduces a visualization of stochastic dominance. Analyzing permutation problems, specifically the Quadratic Assignment Problem, the thesis reveals insights into the influence of distances on optimization algorithms. Additionally, a novel multi-domain problem analysis method using hyper-heuristics is proposed, and a general early stopping criterion, GESP, is introduced to address lengthy evaluation times. The findings contribute to the enhancement of optimization algorithms, offering valuable insights for diverse problem domains.