| Paper Details: | Downloads: 717 |
| Serial Number: | P1121213150
|
| Title: | Designing of an Efficient Classifier using Hierarchical Reinforcement Learning
|
| Authors: | Moumita Saha, Jaya Sil, Nandita Sengupta
|
| Abstract: | Large dimensional real life dataset often consists of vague and redundant information, creating difficulty in building an efficient classifier. In the early part of this work, fuzzy-rough set and genetic algorithm (GA) were applied on continuous domain to select important features sufficient to classify the dataset, called reducts. The aim of the paper is to generate fuzzy rule set with optimum variation in the range of linguistic labels representing antecedent of the rules. Hierarchical Q-Learning method consisting of two levels, are applied to achieve the goal. In the first level of learning, optimized variation is achieved with respect to each reduct and in the second level, learning is applied to optimize the variation with respect to each attribute of the reduct. The performance of the classifiers are evaluated before and after learning, demonstrating improvement in classification accuracy by imparting training.
|
| Keywords: | Q Learning, Hierarchical Q Learning
|
| Journal/Conference: | ICGST International Conference on Computer Science and Engineering, CSE-Dubai-12
|
| Volume: |
|
| Issue: |
|
| Submission Date: | 3/30/2012 12:00:00 AM
|
| Review Date: | 6/5/2012 8:32:50 PM
|
| Publishing Date: | 7/16/2012 12:00:00 AM
|
| Article Downloads: | 717
|
| Download: |
|