Active Data Partitioning for Building Mixture Models
Suk-Joon Kim, Byoung‐Tak Zhang · 1998
This paper introduces two data partitioning methods for building mixtures of several neural networks. The methods are based on active learning with two different selection measures. One is the redundant data selection (RDS) method whiCh chooses examples with less error, and the other is the critical data selection (CDS) method which chooses examples with larger error. The partitioned data sets are used to train the experts which are then combined by a weighted majority algorithm to produce final outputs. Experiments have been performed on two data sets from the UCI machine learning database. The results show that CDS outperforms both RDS and random selection in generalization ability. We also suggest a promising way to use the data subsets partitioned by RDS.