http://www.cs.ubc.ca/labs/beta/Projects/SATzilla/ http://www.csail.mit.edu/events/eventcalendar/calendar.php?show=event&id=2784
Automated Algorithm Configuration and Selection: Enabling Technologies for Building Better Algorithms Speaker: Frank Hutter, University of British Columbia, Vancouver Date: Monday, December 6 2010 Time: 5:00PM to 6:00PM Refreshments: 4:45PM Location: 32-D463 - Star Conference Room Host: Vijay Ganesh, MIT-CSAIL Contact: Mary McDavitt, 617-253-9620, [email protected] Relevant URL: Abstract: Algorithms for solving difficult computational problems play a key role in many applications, including scheduling, time-tabling, resource allocation,production planning and optimization, computer- aided design, and software verification. In many cases, provably efficient algorithms are unlikely to exist, and heuristic methods are the key to solving these problems effectively. However, the design of effective heuristic algorithms is a difficult task that requires substantial expertise and time. Traditionally, it involves an iterative, manual process, in which the designer gradually introduces or modifies components or mechanisms whose empirical performance is then tested on one or more sets of benchmark problems. In this talk, we describe our research on fully formalized methods that automate the most tedious and time-consuming part of this algorithm design process. In particular, we discuss two automated algorithm configuration frameworks, which aim at identifying the combination of algorithm components with the best empirical performance for a given application domain. We also describe our work on algorithm selection, which aims at selecting the best algorithm on a per-instance basis, as well as a recent extension for selecting an algorithm's best components on a per-instance basis. We illustrate the power of these fully automated methods on examples from SAT-based verification and mixed integer programming. Without the need for domain knowledge or human time, in several cases they sped up hand- crafted high-performance algorithms by orders of magnitude, thereby substantially advancing the state of the art in solving a broad range of problems. Based on these results, we believe that automated methods such as the ones we present will play an increasingly crucial role in the design of high-performance algorithms and will be widely used in academia and industry. Based on joint work with Holger Hoos, Kevin Leyton-Brown, and many others Bio: Frank Hutter is a Postdoctoral Research Fellow at the Computer Science Department of the University of British Columbia in Vancouver, Canada, where he works with Profs. Holger Hoos and Kevin Leyton-Brown. His research concentrates on the use of machine learning and optimization to improve algorithms for solving NP-hard problems -- You received this message because you are subscribed to the Google Groups "Algorithm Geeks" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/algogeeks?hl=en.
