PPSN 2012

Joint Workshop on Automated Selection and Tuning of Algorithms (ASTA)

The steadily growing supply of new optimization methods makes the algorithm selection problem an increasingly pressing and challenging task, both in continuous as well as in discrete combinatorial optimization. Therefore, choosing and tuning a suitable optimization algorithm for a given instance of an optimization problem is a crucial issue and should be supported by automated tools based on problem characteristics.

The aim of this workshop is to collect research concepts and bring together researchers of interdisciplinary areas such as computer science, artificial intelligence, statistics, machine learning and optimization in order to interactively discuss the current state-of-the art and most important research topics to be addressed in the near future.

This joint workshop is part of the workshop programme of the PPSN 2012 conference. It contains two related parts with slightly different foci:


Workshop Program, Slides and Related Material



Part A: Continuous Search Spaces -- Focus on Algorithm Selection

The steadily growing supply of new optimization methods makes the algorithm selection problem (ASP) an increasingly pressing and challenging task, e.g. for real-world black-box optimization problems and machine learning tasks. Therefore, choosing a suitable optimization algorithm for a given instance of an optimization problem should become an automated process.

The aim of this workshop is to collect research concepts and bring together researchers of interdisciplinary areas such as computer science, artificial intelligence, statistics, machine learning and optimization in order to interactively discuss the current state-of-the art and most important research topics to be addressed in the near future.

Relevant topics for this workshop are with focus on continuous optimization:

  • Methods for Algorithm Selection in Algorithm Portfolios
  • Benchmarking Theory
  • Design of Benchmarking Competitions
  • Exploratory Landscape Analysis
  • Feature-Based Prediction of Algorithm Performance
  • Fitness Landscape Analysis
  • Algorithm Performance Assessment Methods
  • Novel Visualization Techniques for Comparing Algorithm Performance
  • Test Problem Design
  • Machine Learning Techniques

Workshop Format and Submissions

The workshop aims at interactive talks and discussions integrating original contributions as well as brief talks about possible research directions and open questions. Feedback of the participants w.r.t. current research will foster publications in progress and provide a basis for additional collaborative research projects.

A panel discussion with invited speakers will complete the discussions on key prospective research directions. It will be placed after both parts of the workshop and will include speakers of both research directions to further foster the interdisciplinarity of the workshop.

Organizing Committee

Dr. Heike Trautmann
Department of Computational Statistics
Statistics Faculty
TU Dortmund
Germany
Tel.: 0049 231 755-3123
Fax.: 0049 231 755-4387
e-mail: trautmann@statistik.tu-dortmund.de

Heike Trautmann is a postdoctoral researcher at the Statistics Department, TU Dortmund, Germany. She studied Statistics, and after working in a consulting company for two years, she joined the Graduate School of Production Engineering and Logistics at the TU Dortmund and received her PhD in 2004. Her current research activities are focused on multiobjective (evolutionary) optimisation – in particular preference incorporation, performance assessment and stopping criteria – as well as benchmarking concepts and Exploratory Landscape Analysis (ELA). She was involved in organizing the special session “Designing Evolutionary Processes” at CEC 2010.

Mike Preuss
Chair of Algorithm Engineering
TU Dortmund
Germany
Tel.: 0049 231 755-7708
Fax.: 0049 231 755-7740
e-mail: mike.preuss@tu-dortmund.de

Mike Preuss is Research Associate at the Computer Science Department, TU Dortmund, Germany, where he also received his Diploma degree in 1998. His research interests focus on the field of evolutionary algorithms for real-valued problems, namely on multimodal and multiobjective niching and the experimental methodology for (non-deterministic) optimization algorithms. He is currently working on the adaptability and applicability of computational intelligence techniques for various engineering domains and computer games, pushing forward modern approaches of experimental analysis as the Exploratory Landscape Analysis (ELA) and innovative uses of surrogate models. He was involved in founding the EvoGames track at Evo* and the Digital Entertainment Technologies and Arts (DETA) track at GECCO.

M.Sc. Olaf Mersmann
Department of Computational Statistics
Statistics Faculty
TU Dortmund
Germany
Tel.: 0049 231 755-4355
Fax.: 0049 231 755-4387
e-mail: olafm@statistik.tu-dortmund.de

Olaf Mersmann is a Research Associate at the Statistics Department of the TU Dortmund. He recieved his B.Sc. and M.Sc. in Data Science from the TU Dortmund and is currently pursuing a PhD in Statistics with work based on his Bachelor Thesis on the design and analysis of benchmark experiments. Part of this work has been presented at conferences and is currently under review for publication. Using statistical and machine learning methods on large benchmark databases to gain insight into the structure of the algorithm choice problem is one of his current research interests. He is currently involved in the organization of the BBOB Workshop at the forthcoming GECCO conference.

M.Sc. Bernd Bischl
Department of Computational Statistics
Statistics Faculty
TU Dortmund
Germany
Tel.: 0049 231 755-4355
Fax.: 0049 231 755-4387
e-mail: bischl@statistik.tu-dortmund.de

Bernd Bischl is currently a Research Associate of the Research Training Group “Statistical Modelling” at the Statistics Faculty, TU Dortmund, Germany. He studied computer science in Hamburg (Germany) and data science in Dortmund (Germany) where he received his M.Sc. degree in 2009. His research interests focus mainly on machine learning, especially model selection and optimizing machine learning systems. But he is also interested in using machine learning techniques to improve optimization methods.

Part B: Discrete Search Spaces -- Focus on Parameter Selection

Scaling Behaviours of Landscapes, Parameters and Algorithms

All too often heuristics and meta-heuristics for discrete combinatorial optimisation problems require significant parameter tuning to work most effectively. Often this tuning is performed without any a priori knowledge as to how good values of parameters might depend on features of the problem. This lack of knowledge can lead to lot of computational effort and also has the danger of being limited to only problem instances that are similar to those that have been seen before. The aim of the workshop is to develop methods to give deeper insight into problem classes, and how to obtain and exploit structural information. In particular, we often would like to be able to tune parameters using small instances (for speed) but then adjust so as to be able to run on large instances. This will require some theory of how to extrapolate tuning outside of the size or features of the training set. An analogy is the difference between non-parametric and parametric statistics; the former does not assume any underlying probability distribution and the latter can (for example) assume a Gaussian. Naturally, the latter might give stronger results and with smaller sized samples. Hence, to distinguish this from standard parameter tuning, we might call this “Parametric Parameter Tuning”. Of course, this is a challenging problem; but we hope to be able to discuss any existing work and how the community might meet the challenge.

Related to this is the common and natural belief that the semantic properties of the landscapes will be reflected in the performances of algorithms. A subsequent underlying assumption, or hypothesis, if the landscape has a particular functional dependence on features of the instance, then such functional dependencies are also likely to play a key role in understanding the behaviour of heuristic algorithms, and so merit investigation. We are particularly interested the area of phase transitions; when particular semantic properties display phases of 'almost always true' and 'almost never true'. Statistical methods can then reveal some appropriate parameters to describe the locations of such phases, and we expect that this will also influence the understanding, design and tuning of algorithms. This is exemplified by the work in the artificial intelligence and statistical physics communities on propositional satisfiability and graph colouring, and that has led to deeper understanding of algorithms, and development of new ones. One of the goals of the workshop is to look into phase transition theory with a view to potential applications to traditional PPSN problems.

Workshop Format

All talks will be encouraged to be interactive, with plenty of time for questions and discussion. We intend that the final discussion session will be loosely formatted around a panel and designed to

  1. give and receive constructive comments about the presented work,
  2. obtain wide range of views,
  3. highlight any differences of philosophy
  4. raise awareness of other lines of work in the area

Organizing Committee

Dr Ender Özcan
School of Computer Science
University of Nottingham
Jubilee Campus, Wollaton Road
Nottingham, NG8 1BB, UK
Tel: +44(0) 115 95 15544
Fax:+44(0) 115 9514254
http://www.cs.nott.ac.uk/~exo/
exo@cs.nott.ac.uk

Ender Özcan is a lecturer in Operational Research and Computer Science with the Automated Scheduling, Optimisation and Planning (ASAP) research group in the School of Computer Science at the University of Nottingham, UK. He received his PhD from the Department of Computer and Information Science at Syracuse University, NY, USA in 1998. He worked as a lecturer in the Department of Computer Engineering at Yeditepe University, Istanbul, Turkey from 1998-2007. He established and led the ARTIficial Intelligence research group from 2002. He served as the Deputy Head of the Department from 2004-2007. Dr Özcan joined the ASAP group as a senior research fellow in 2008. He has been serving as an executive committee member for the LANCS initiative, which is one of the largest Science and Innovation Rewards given by EPSRC (Engineering and Physical Sciences Research Council, UK). His main research interests include intelligent decision support systems, search and optimisation using heuristics, hyper-heuristics, meta-heuristics, hybrid approaches and their theoretical foundations, and their applications to the real world and theoretical problems.

Dr Andrew J. Parkes
School of Computer Science
University of Nottingham
Jubilee Campus, Wollaton Road
Nottingham, NG8 1BB, UK
Tel: +44(0) 115 95 14210
Fax:+44(0) 115 9514254
http://www.cs.nott.ac.uk/~ajp/
ajp@cs.nott.ac.uk

Andrew Parkes obtained a Ph.D. in theoretical physics in 1984 from King's College of the University of London. He then held research positions at the University of Southampton, CERN (the particle physics accelerator and research laboratory at Geneva, Switzerland), the University of California at Davis, USA, and the ETH in Zurich, Switzerland. In 1993 he changed from Physics to Artificial Intelligence and Computer Science. In 1999 he obtained a Ph.D., in Computer Science from the University of Oregon, USA. He became a faculty member of the Computational Intelligence Research Laboratory at the University of Oregon working on topics in combinatorial optimisation. In 2005 he joined the Automated Scheduling, Optimisation and Planning Research Group (ASAP) within the School of Computer Science of the University of Nottingham where he researched the management and planning of teaching space facilities within educational institutions. Since 2008 he has been a Lecturer in Operational Research and Computer Science at the University of Nottingham with particular responsibility to support the recent EPSRC-funded Science and Innovation award for “The LANCS Initiative: An initiative to build and maintain national research capacity in foundational Operational Research.” He is a member of the LANCS executive committee and also coordinating a research cluster into the understanding of heuristics. He was on the organising committee for the 2007 International Timetabling competition. He is a co- inventor on two patents and has many refereed international publications.

 
Last modified: 2015-09-11 13:56 (external edit)
DokuWikiRSS-Feed