A Rapid Graph-Based Method for Arbitrary Transformation-Invariant Pattern Classification
Abstract
We present a graph-based method for rapid, accurate search through prototypes for transformation-invariant pattern classifica(cid:173) tion. Our method has in theory the same recognition accuracy as other recent methods based on ''tangent distance" [Simard et al., 1994], since it uses the same categorization rule. Nevertheless ours is significantly faster during classification because far fewer tan(cid:173) gent distances need be computed. Crucial to the success of our system are 1) a novel graph architecture in which transformation constraints and geometric relationships among prototypes are en(cid:173) coded during learning, and 2) an improved graph search criterion, used during classification. These architectural insights are applica(cid:173) ble to a wide range of problem domains. Here we demonstrate that on a handwriting recognition task, a basic implementation of our system requires less than half the computation of the Euclidean sorting method.
Cite
Text
Sperduti and Stork. "A Rapid Graph-Based Method for Arbitrary Transformation-Invariant Pattern Classification." Neural Information Processing Systems, 1994.Markdown
[Sperduti and Stork. "A Rapid Graph-Based Method for Arbitrary Transformation-Invariant Pattern Classification." Neural Information Processing Systems, 1994.](https://mlanthology.org/neurips/1994/sperduti1994neurips-rapid/)BibTeX
@inproceedings{sperduti1994neurips-rapid,
title = {{A Rapid Graph-Based Method for Arbitrary Transformation-Invariant Pattern Classification}},
author = {Sperduti, Alessandro and Stork, David G.},
booktitle = {Neural Information Processing Systems},
year = {1994},
pages = {665-672},
url = {https://mlanthology.org/neurips/1994/sperduti1994neurips-rapid/}
}