NIPS'97 Workshop

Neural Models of Concept Learning

Neuronal models of lexical acquisition and semantic processing have been addressed in the context of connectionist models of categorization, including self-organizing maps, constraint satisfaction networks and multilayer feedforward networks. Recently, a number of new suggestions and proposals have been forthcoming which put research on lexical meaning in context with concept learning, language development and statistical induction. These approaches relate to the disciplines of

• cognitive psychology of concept learning,
• statistical lexical acquisition,
• emergentist approaches to language development.

It seems that a theory of conceptual semantics is forthcoming which pushes beyond the limits of connectionist models by incorporating strands from dynamic system theory (attractor networks, catastrophe theory) and statistical induction (Kolmogorov complexity, MDL learning).

The aim of the workshop is to provide a forum for an interchange of methodologies, research directions and theoretical discussion. The format will consist of introductory statements from panelists, group discussion and individual position paper presentations.

Confirmed participants:

• Chris Browne, Imperial College, London
• Curt Burgess, University of Claifornia, Riverside
• Daniel Clouse, University of California, San Diego
• Eliana Colunga, Indiana University
• Georg Dorffner, University of Vienna
• Michael Gasser, Indiana University
• Mark Gluck, Rutgers University
• Catherine Harris, Boston University
• Michel de Heaulme, CHU Pitie-Salpetriere, Paris
• Ping Li, University of Richmond
• Kevin Lund, University of California, Riverside
• Risto Miikkulainen, University of Texas, Austin
• Terry Regier, University of Chicago
• Gabriele Scheler, Munich University of Technology
• Philipe Schyns, University of Glasgow
• Joshua Tenenbaum, MIT
• Chris Thornton, University of Sussex

Additonal contributions by (see abstracts):

• T. Honkela, Helsinki Univ. of Technology
• Maartje Raijmakers, University of Amsterdam

Schedule

Abstracts

Discussion topics and questions

Workshop Organization:

Gabriele Scheler, Munich University of Technology and ICSI, Berkeley
Georg Dorffner, University of Vienna and Austrian Research Institute for Artificial Intelligence

Conference Site:

The workshop will be held in conjunction with NIPS'97 in Breckenridge, Colorado on December 6, 1997. For more information on the workshops, see here.

Format:

We envisage about 20-30 participants. There will be 12 individual paper presentations of 10-15 minutes each, a number of panelists' statements and an open group discussion.

Publication/Dissemination:

Electronic publication of submitted abstracts is being realized here. Further electronic and print publication as a survey paper in Neural Computing Surveys is planned. A book publication by Springer or another major publisher in 1998 based on full papers from workshop participants and additional contributions by other researchers is anticipated.