François Gonard, PhD student at both SystemX and Inria, participated in the ICON Challenge on Algorithm Selection.
He obtained the Honorable Mention, a promising result for his thesis, and for the SystemX ROM project (Model Reduction and Multiphysics Optimization) which aims at improving the design and optimization process, among other challenges.
This ICON Challenge has been organized by the Insight Centre for Data Analytics, University College Cork, as part of the European FET (Future and Emerging Technologies) Program “Open”. This program supports early-stage research on any idea dealing with a new technology and encourages scientists and engineers from multiple disciplines to work together on game-changing science and technology research, expanding well beyond the traditional technological disciplines.
The candidates who applied to this challenge were asked to use their machine learning skills to build an adaptive system that guesses for each problem instance to solve the algorithms that fit the most – i.e. the fastest ones. The challenge received a total of 8 submissions from 4 different groups of researchers comprising 15 people. All submissions were evaluated on the ASlib (a benchmark library for algorithm selection) scenarios.
The Honorable Mention was granted due to the high performance system that François developed. His submitted work ASAP_RF rivaled state-of-the-art systems that have already been in development for several years, beating all of them in some cases. ASAP_RF was in addition the most balanced, meaning the results varied the least during tests, compared to some systems that tended to fluctuate depending on the situation.
In addition of receiving this Honorable Mention, his system can already be considered as industrial grade material, and be used as a solid basis for the system developed in the ROM project to support the design and optimization process. As a PhD student whose thesis is tied into the problematic of Algorithm Selection, it is also a strong addition to his work. François Gonard will be participating in the COSEAL Workshop, and will be presenting his submission ASAP_RF.
Challenge results available here: http://www.cs.ubc.ca/~larsko/aschallenge-results.pdf