Insitu Learning in Multinet Systems Surrey在多网系统现场学习萨里.pptVIP

Neural Computing Group Department of Computing University of Surrey Classification and Multi-nets Multiple classifiers: Can give improved classification [6, 16] Ensemble: redundancy Modular: no redundancy, simpler classifiers Multi-net systems [12]: Systems consisting of multiple neural networks Parallel and sequential combinations Mixture-of-experts (ME) [5]: modular, in-situ In-situ Learning ME learns how to combine networks in parallel Simultaneous training of networks and combination Existing techniques focus on pre-training Ensembles Can in-situ learning in different types of multi-net s