Description

We have constructed a linear discriminator for hand-printed character recognition that uses a (binary) vector of 1,500 features based on an equidistributed collection of products of pixel pairs. This classifier is competitive with other techniques, but faster to train and to run for classification. However, the 1,500-member feature set clearly contains many redundant (overlapping or useless) members, anda significantly smaller set would be very desirable (e.g., for faster training, a faster and smaller application program, and a smaller system suitable for hardware implementation). A system using the small set of features should also be better at generalization, since fewer features are less likely to allow a system to "memorize noise in the training data." Several approaches to using a genetic algorithm to search for effective small subsets of features have been tried, and we have successfully derived a 300-element set of features and built a classifier whose performance is as good on our training and testing set as the system using the full set.

Date of creation, presentation, or exhibit

3-21-1995

Comments

Artificial Neural Networks and Genetic Algorithm (1993) 101-106 "Genetic algorithm selection of features for hand-printed character identification," Presented at The International Conference on Artificial Neural Networks and Genetic Algorithms "ANNGA 93". Held in Innsbruck, Austria, April 1993. Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works in February 2014.

Document Type

Conference Proceeding

Department, Program, or Center

Chester F. Carlson Center for Imaging Science (COS)

Campus

RIT – Main Campus

Share

COinS