New Boosting Approach Using Probabilistic Results Expression
| Autor: | Won Don Lee, A. Kolesnikova, Dong-Hun Seo |
|---|---|
| Rok vydání: | 2008 |
| Předmět: |
Probabilistic classification
Training set Boosting (machine learning) business.industry Computer science Probabilistic logic Decision tree Pattern recognition Machine learning computer.software_genre Statistical classification Incremental learning Artificial intelligence business computer Classifier (UML) |
| Zdroj: | International Symposium on Computer Science and its Applications. |
| DOI: | 10.1109/csa.2008.63 |
| Popis: | Classification, which is the task of assigning object (event) to one of several predefined classes, is an important problem in the field of machine learning and data mining. Boosting based technique use deterministic results of weak classifiers to compound them to a strong classifier while classifier can give class distribution as result. It involves losses of information. This problem can be solved by using probabilistic idea. In this case class distribution is used to compound classifiers. Probabilistic results compounding is presented in this paper applying to incremental learning. Probabilistic results expression is easily realized using extended data expression. Results of experiments show power when compare to Learn++, an incremental ensemble-based algorithm. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|