Formation of Probabilistic Concept Hierarchies
Upon moving to UCI in 1984, I began a long-term collaboration with
Douglas Fisher, which led initially to analyses of methods for
clustering and taxonomy formation. When Fisher developed Cobweb,
an incremental unsupervised algorithm that formed hierarchies of
probabilistic concept descriptions, I decided it would serve as an
excellent base for Icarus, an architecture
for physical agents that I was developing with other colleagues.
Over the next few years, our group developed a number of extensions
to Cobweb. These included John Gennari's Classit (which handled
numeric attributes and supported an attention mechanism), Kevin
Thompson's Labyrinth (which dealt with concept formation over
composite, structured objects), John Allen's Daedalus (which
stored probabilistic summaries of planning decisions), and
Maeander (which formed probabilistic hierarchies of motor skills).
After moving to NASA Ames, our group continued this work and
attempted to integrate the various components in the Icarus
architecture. Problems with order effects led to Arachne, a
variant of Cobweb developed with Kate McKusick, and a growing
interest in the naive Bayesian classifier.
Our joint research on Cobweb and its relatives wound down in 1992,
when I left NASA for greener pastures.
The list below does not include all publications related to
Cobweb. Rather, it focuses on those papers that explore the
basic algorithm and its uses in classification and prediction,
rather than extensions to more complex problems.
Iba, W. F., & Langley, P. (2011). Cobweb models of categorization
and probabilistic concept formation. In E. M. Pothos & A. J. Willis
(Eds.), Formal approaches in categorization. Cambridge:
Cambridge University Press.
Iba, W., & Langley, P. (2001).
Unsupervised learning of probabilistic concept hierarchies.
In G. Paliouras, V. Karkaletsis, & C. D. Spyropoulos (Eds).,
Machine learning and its applications. Berlin: Springer.
Thompson, K., & Langley, P. (1991).
Concept formation in structured domains.
In D. H. Fisher, M. J. Pazzani, & P. Langley (Eds.),
Concept formation: Knowledge and experience in unsupervised
learning. San Mateo, CA: Morgan Kaufmann.
McKusick, K. B., & Langley, P. (1991).
Constraints on tree structure in concept formation.
Proceedings of the Twelfth International Joint Conference on
Artificial Intelligence (pp. 810-816). Sydney: Morgan Kaufmann.
Thompson, K., Langley, P., & Iba, W. F. (1991).
Using background knowledge in concept formation.
Proceedings of the Eighth International Workshop on Machine
Learning (pp. 554-558). Evanston, IL: Morgan Kaufmann.
Fisher, D. H., & Langley, P. (1990).
The structure and formation of natural categories.
In G. H. Bower (Ed.), The psychology of learning and motivation:
Advances in Research and Theory (Vol. 26).
Cambridge, MA: Academic Press.
Billman, D., Fisher, D., Gluck, M., Langley, P., & Pazzani, M. (1990).
Computational models of category learning.
Proceedings of the Twelfth Conference of the Cognitive Science
Society (pp. 989-996). Cambridge, MA: Lawrence Erlbaum.
Gennari, J. H., Langley, P., & Fisher, D. H. (1989).
Models of incremental concept formation.
Artificial Intelligence, 40, 11-61.
Fisher, D., & Langley, P. (1986).
Methods of conceptual clustering and their relation to numerical
In W. Gale (Ed.), Artificial intelligence and statistics.
Reading, MA: Addison Wesley.
Easterlin, J. D., & Langley, P. (1985).
A framework for concept formation.
Proceedings of the Seventh Conference of the Cognitive Science
Society (pp. 267-271). Irvine, CA.
Fisher, D., & Langley, P. (1985).
Approaches to conceptual clustering.
Proceedings of the Ninth International Joint Conference on
Artificial Intelligence (pp. 691-697).
Los Angeles, CA: Morgan Kaufmann.
For more information, send electronic mail to