论文信息 - Evaluation of Adaptive Mixtures of Competing Experts

Evaluation of Adaptive Mixtures of Competing Experts

We compare the performance of the modular architecture, composed of competing expert networks, suggested by Jacobs, Jordan, Nowlan and Hinton (1991) to the performance of a single back-propagation network on a complex, but low-dimensional, vowel recognition task. Simulations reveal that this system is capable of uncovering interesting decompositions in a complex task. The type of decomposition is strongly influenced by the nature of the input to the gating network that decides which expert to use for each case. The modular architecture also exhibits consistently better generalization on many variations of the task.

Geoffrey E. Hinton | Steven J. Nowlan | S. Nowlan

[1] G. E. Peterson,et al. Control Methods Used in a Study of the Vowels , 1951 .

[2] L. Gerstman. Classification of self-normalized vowels , 1968 .

[3] Michael I. Jordan,et al. A Competitive Modular Connectionist Architecture , 1990, NIPS.

[4] Steven J. Nowlan,et al. Soft competitive adaptation: neural network learning algorithms based on fitting statistical mixtures , 1991 .

[5] Michael I. Jordan,et al. Task Decomposition Through Competition in a Modular Connectionist Architecture: The What and Where Vision Tasks , 1990, Cogn. Sci..

[6] Geoffrey E. Hinton,et al. Adaptive Mixtures of Local Experts , 1991, Neural Computation.