Reservoir Computing with a Small-World Network for Discriminating Two Sequential Stimuli
Recently, reservoir network was used for simulating the sequential stimuli discrimination process of monkeys. To deal with the inefficient memory problem of a randomly connected network, a winner-take-all subnetwork was used. In this study, we show that a network with the small-world property makes the WTA subnetwork unnecessary. Using the reinforcement learning in the output layer only, the proposed network successfully learns to accomplish the same discrimination task. In addition, the model neurons exhibit heterogeneous firing properties, which is consistent with the physiological data.
KeywordsReservoir computing Small-world network Decision making Reinforcement learning
This work was supported in part by the National Basic Research Program (973 Program) of China under Grant 2013CB329403, in part by the National Natural Science Foundation of China under Grant 91420201, Grant 61332007, and Grant 61621136008, and in part by the German Research Foundation (DFG) under Grant TRR-169.
- 1.Baddeley, A.D.: Working Memory. Oxfoid Psychology Series, vol. 11. Oxford University Press, New York (1986)Google Scholar
- 5.Jaeger, H.: The “echo state” approach to analysing and training recurrent neural networks-with an erratum note. Bonn Ger.: Ger. Natl. Res. Center Inf. Technol. GMD Tech. Rep. 148, 34 (2001)Google Scholar