Portal

Causality workbench - Isabelle Guyon

• Report rights infringement
• published: 16 Dec 2013
• views: 400

This talk is part 6 of the Workshop on Case Studies of Causal Discovery with Model Search, held on October 25-27, 2013, at Carnegie Mellon University. Speaker: Isabelle Guyon Title: Causality workbench Abstract: Challenges have recently proved a great stimulus for research in machine learning, pattern recognition, and robotics. Attracting hundreds of participants and the attention of a broad audience of specialists as well as sometimes the general public, these events have been important in several respects: (1) pushing the state-of-the art, (2) identifying techniques which really work, (3) attracting new researchers, (4) raising the standards of research, (5) giving the opportunity to non-established researchers to make themselves rapidly known. Since 2007, we have been organizing challenges in causal discovery. To that end, we have set up a dedicated challenge platform, the Causality Workbench. Via a web portal (http: //clopinet.com/causality), the Causality Workbench provides a number of resources, including a repository of datasets, models, and software packages, and a virtual laboratory allowing users to benchmark causal discovery algorithms by performing virtual experiments to study artificial causal systems. Our effort has been gaining momentum with the organization of three challenges, attracting in total several hundreds of participants. Our challenge websites, which remain open for post-challenge submissions, are constantly in use by students and have been used in practical work in our own classes and those of other professors throughout the world. We take great care of giving to the participants opportunities publish in reputable conferences proceedings or journals like JMLR. We think of challenges as a means of carrying out research in causal discovery by focusing the mental energy of brilliant researchers around the world. But, what makes a good challenge that provides conclusive results having an important impact? This presentation will review the main findings of our past challenges and look upon them with a critical eye to identify strength and weaknesses and new directions. Link to slides: TBA Link to workshop homepage: http://www.hss.cmu.edu/philosophy/casestudiesworkshop.php Workshop description: Computer scientists, statisticians, and philosophers have created a precise mathematical framework for representing causal systems called "Graphical Causal Models." This framework has supported the rigorous description of causal model spaces and the notion of empirical indistinguishability/equivalence within such spaces, which has in turn enabled computer scientists to develop asymptotically reliable model search algorithms for efficiently searching these spaces. The conditions under which these methods are practically useful in applied science was the topic of this workshop. The workshop brought together scholars from genetics, biology, economics, fMRI-based cognitive neuroscience, climate research, education research, and several other disciplines, all of whom have successfully applied computerized search for causal models toward a scientifically challenging problem. The goals for the workshop were: (1) to identify strategies for applying causal model search to diverse domain-specific scientific questions; (2) to identify and discuss methodological challenges that arise when applying causal model search to real-world scientific problems; and (3) to take concrete steps toward creating an interdisciplinary community of researchers interested in applied causal model search. We welcomed junior scholars and graduate students, and we hosted a free introductory tutorial on model search the first morning of the workshop. Funding for this workshop was provided by the National Science Foundation, grant #SES1156001, and by the Center for Formal Epistemology at Carnegie Mellon University. Thanks to their support, we were able both to host the workshop, and also record it, making it accessible to a larger audience.