Incremental Art: A Neural Network System for Recognition by Incremental Feature Extraction
OA Version
Citation
Abstract
Abstract Incremental ART extends adaptive resonance theory (ART) by incorporating mechanisms for efficient recognition through incremental feature extraction. The system achieves efficient confident prediction through the controlled acquisition of only those features necessary to discriminate an input pattern. These capabilities are achieved through three modifications to the fuzzy ART system: (1) A partial feature vector complement coding rule extends fuzzy ART logic to allow recognition based on partial feature vectors. (2) The addition of a F2 decision criterion to measure ART predictive confidence. (3) An incremental feature extraction layer computes the next feature to extract based on a measure of predictive value. Our system is demonstrated on a face recognition problem but has general applicability as a machine vision solution and as model for studying scanning patterns.
Description
License
Copyright 1994 Boston University. Permission to copy without fee all or part of this material is granted provided that: 1. The copies are not made or distributed for direct commercial advantage; 2. the report title, author, document number, and release date appear, and notice is given that copying is by permission of BOSTON UNIVERSITY TRUSTEES. To copy otherwise, or to republish, requires a fee and / or special permission.