TWiki> Isaim2012 Web>News>SpecialSessions (revision 13)EditAttach

Special Sessions

ISAIM 2012 will feature several special sessions:

  • Computational Social Choice
  • Boolean and pseudo-Boolean Functions
  • Causal Learning from Complex Data Structures
  • Arrangements for additional special sessions are currently being made
Proposals for additional Special Sessions may be sent to the Organizers by October 1, 2011.

Computational Social Choice

Organizer: Judy Goldsmith; Publicity Chair: Nicholas Mattei

Computational Social Choice (ComSoc) is an emerging and rapidly evolving subfield of artificial intelligence. ComSoc is focused on how agents make collective decisions. Social Choice, an established research field at the intersection of mathematics and political science, has long studied the implications of group decisions in human systems. With the growth of multi-agent systems research in the AI community it has become necessary to investigate how agents can work together and make group decisions. ComSoc and social choice are related by two main bridges: bringing a computational perspective to decision systems already in use and/or studied by social choice, and bringing systems and processes developed through years of social choice research to bear on multi-agent systems. Research areas that fall under the ComSoc include (but are not limited to) voting and election systems; fair division algorithms; coalition formation and management; judgement aggregation and belief merging; and stable matching problems.

Boolean and pseudo-Boolean Functions

Organizers: Endre Boros and Yves Crama

Boolean and pseudo-Boolean functions are pervasive today in all areas of mathematics, computer science, operations research, various sciences and engineering. An ever increasing number and areas of applications demand new results from both structural and algorithmic points of views. The special sessions aim at bringing together researchers from all walks of science to discuss the latest results and the most important open problems.

Causal Learning from Complex Data Structures

Organizer: David Danks, Carnegie Mellon

Over the past twenty years, a range of algorithms have been developed for learning causal structure from relatively straightforward data: a single dataset with fully-observed variables (or at most, missing at random values) where all individuals in the dataset are part of the same experimental design (if any). For continuous, real-valued variables, these algorithms typically further assume that any dependencies are linear. Different algorithms might weaken one or another of these assumptions - for example, one might allow for the possibility of unobserved common causes - but the basic focus on learning from a single dataset measuring a causally homogeneous population is almost universal.

In contrast, real-world science often requires integrating information from multiple datasets that need not measure the same variables, nor use the same experimental design, nor measure linear relationships. In the past five years, a number of different algorithms have emerged that learn causal structure in these more complex situations. This special session will bring together some of the key researchers in the development of those algorithms to discuss the latest results, natural extensions, and possibilities for integration of the different techniques.

Edit | Attach | Print version | History: r17 | r15 < r14 < r13 < r12 | Backlinks | Raw View | Raw edit | More topic actions...
Topic revision: r13 - 2011-08-30 - 16:08:21 - Main.ddioch2
Copyright 2016 The Board of Trustees
of the University of
Helping Women Faculty Advance
Funded by NSF