TRESTLE: Incremental Learning in Structured Domains using Partial Matching and Categorization
Proceedings of the Third Annual Conference on Advances in Cognitive Systems
2015
Abstract
We present TRESTLE, an incremental algorithm for probabilistic concept formation in structured domains that builds on prior concept learning research. TRESTLE works by creating a hierarchical categorization tree that can be used to predict missing attribute values and cluster sets of examples into conceptually meaningful groups. It is able to update its knowledge by partially matching novel structures and sorting them into its categorization tree. The algorithm supports mixed-data representations, including nominal, numeric, relational, and component attributes. We evaluate the algorithm’s performance on prediction and categorization tasks and show preliminary evidence that this new categorization model is competitive with non-incremental approaches and more closely approximates human performance.
BibTeX
@inproceedings{maclellan-acs-2015,
title = {TRESTLE: Incremental Learning in Structured Domains using Partial Matching and Categorization},
author = {MacLellan, Christopher J. and Harpstead, Erik and Aleven, Vincent and Koedinger, Kenneth R.},
booktitle = {Proceedings of the Third Annual Conference on Advances in Cognitive Systems},
year = {2015},
}