by Wojciech Jaśkowski, Krzysztof Krawiec, Bartosz Wieloch
Abstract:
We propose a learning algorithm that reuses knowledge acquired inpast learning sessions to improve its performance on a new learningtask. The method concerns visual learning and uses genetic programmingto represent hypotheses, each of them being a procedure that processesvisual primitives derived from the training images. The process ofrecognition is generative, i.e., a procedure is supposed to restorethe shape of the processed object by drawing its reproduction ona separate canvas. This basic method is extended with a knowledge reuse mechanism that allows learners to import genetic material from hypotheses that evolved for the other decision classes (object classes).We compare both methods on a task of handwritten character recognition,and conclude that knowledge reuse leads to significant improvementof classification accuracy and reduces the risk of overfitting.
Reference:
Evolutionary Learning with Cross-Class Knowledge Reuse for Handwritten Character Recognition (Wojciech Jaśkowski, Krzysztof Krawiec, Bartosz Wieloch), In Proceedings of Planning to Learn Workshop, PlanLearn’07 (Pavel Brazdil, Abraham Bernstain, eds.), 2007.
Bibtex Entry:
@InProceedings{Jaskowski2007evolutionary, Title = {Evolutionary Learning with Cross-Class Knowledge Reuse for Handwritten Character Recognition}, Author = {Wojciech Jaśkowski and Krzysztof Krawiec and Bartosz Wieloch}, Booktitle = {Proceedings of Planning to Learn Workshop, PlanLearn'07}, Year = {2007}, Editor = {Pavel Brazdil and Abraham Bernstain}, Pages = {21--30 }, Abstract = { We propose a learning algorithm that reuses knowledge acquired inpast learning sessions to improve its performance on a new learningtask. The method concerns visual learning and uses genetic programmingto represent hypotheses, each of them being a procedure that processesvisual primitives derived from the training images. The process ofrecognition is generative, i.e., a procedure is supposed to restorethe shape of the processed object by drawing its reproduction ona separate canvas. This basic method is extended with a knowledge reuse mechanism that allows learners to import genetic material from hypotheses that evolved for the other decision classes (object classes).We compare both methods on a task of handwritten character recognition,and conclude that knowledge reuse leads to significant improvementof classification accuracy and reduces the risk of overfitting. }, Url = { http://www.ecmlpkdd2007.org/CD/workshops/PlanLearn/WS_PlanLearn_print.pdf } }