Co-evolutionary Multi-task Learning for Modular Pattern Classification
Modularity in the learning process is a means by which effective decision making can be maintained when some of the input features are missing. In this paper, co-evolutionary multi-task learning algorithm is used for pattern classification which is robust to situations when some input features are unavailable during the deployment stage of decision support or control systems. The main feature of the algorithm is the ability to make decisions with some degree of error given misinformation. The results show that the method produces results comparable to non-modular methods while having modular features for dynamic and robust pattern classification.
KeywordsCooperative coevolution Neuroevolution Multi-task learning Modular pattern classification
- 4.Chandra, R., Ong, Y.S., Goh, C.K.: Co-evolutionary multi-task learning for dynamic time series prediction. CoRR abs/1703.01887 (2017). http://arxiv.org/abs/1703.01887
- 15.Lichman, M.: UCI Machine Learning Repository. School of Information and Computer Science, University of California, Irvine (2013). http://archive.ics.uci.edu/ml