Computational Synthesis >From basic building blocks to high level functionality
To be held at Stanford University on March 24-27, 2003 as part of the 2003 AAAI Spring Symposium Series Visit http://www.mae.cornell.edu/ccsl/conf/ PDF version http://www.mae.cornell.edu/ccsl/conf/poster.pdf Organizers: - Hod Lipson (Cornell University) - Erik K. Antonsson (Caltech) - John R. Koza (Stanford University) Computational Synthesis research seeks formal algorithmic procedures that combine low-level building blocks or features to achieve given arbitrary high-level functionality. The main challenge is scaling to high complexities, and the paths of investigation deal with automatic composition of building blocks into useful modules, automatic abstraction of module functionality, and automatic hierarchical reuse of modules. The symposium will focus on domain-independent methods that address modularity, regularity, hierarchy and abstraction in automatic synthesis. Recently there has been a surge of interest in these fundamental issues from three directions: AI researchers interested in scaling discovery processes, engineers interested in fully automated design, and biologists interested in the origin of complexity. Topics of interest include: - Models of bottom-up composition and top-down decomposition - Scalability of composition processes to high complexities - Automatic identification and composition of useful modules - Regularity and hierarchy in composition - Automatic abstraction and encapsulation of modules - Efficient and adaptive representations of design space - Solution-neutral goal specification - Stochastic, game-theoretic and co/evolutionary processes - Machine learning in synthesis - Synthesis as models for discovery in nature and engineering - No free lunch: What can we trade to get open-ended design? Invited Speakers: - Peter J. Bentley (University College London): Natural Design By Computer - Richard E. Michod (University of Arizona): Cooperation and conflict in the evolution of complexity Interested applicants should send a 2-page abstract (in PDF) to [EMAIL PROTECTED] by October 4, 2002 Symposium Format: Diversity of problem-domains is encouraged. The symposium seeks to informally bring together researchers from diverse problem domains to address universal approaches and common issues in automatic synthesis. We welcome technical papers describing proposed or completed research activities; position papers outlining a research agenda or evaluating current ideas and approaches; or suggestions for panel discussions. Particularly welcome are contributions that suggest a generic and domain independent approach, although they might be demonstrated for a particular problem domain. Typical presentations will be short (10 minutes) with most of the time allocated for discussion led by a designated peer. Abstracts will be circulated by mail prior to meeting. *** Student scholarships available ***
