Recurrent Network Models, Reservoir Computing
Reservoir computing (RC) is a general concept for computation and learning on temporal input streams with dynamical systems. Time can either be continuous or discrete. We will here use continuous time for the general definition. An input-driven dynamical system (the reservoir) provides a nonlinearly transformed and temporally integrated representation of the input stream in terms of its internal state. This representation is utilized by a readoutwhich maps internal states to outputs of the system. The output of the readout can be fed back into the reservoir. The reservoir is typically a recurrent neural network, but other dynamical systems have been employed in the RC spirit recently. The readout function is adapted through some learning procedure. Originally, this training was supervised and the reservoir was not adapted. Today, many...
- Boyd S, Chua LO (1985) Fading memory and the problem of approximating nonlinear operators with Volterra series. IEEE Trans Circuits Syst 32:1150–1161Google Scholar
- Buesing L, Schrauwen B, Legenstein R (2010) Connectivity, dynamics, and memory in reservoir computing with binary and analog neurons. Neural Comput 22(5):1272–1311Google Scholar
- Jaeger H (2001) The “echo state” approach to analysing and training recurrent neural networks – with an erratum note. GMD Report 148: German National Research Center for Information TechnologyGoogle Scholar
- Lukoševičius M, Jaeger H (2009) Survey: reservoir computing approaches to recurrent neural network training. Comput Sci Rev 3(3):127–149Google Scholar
- Lukoševičius M, Jaeger H, Schrauwen B (2012) Reservoir computing trends. KI – Künstliche Intelligenz, pp 1–7Google Scholar
- Nikolic D, Haeusler S, Singer W, Maass W (2009) Distributed fading memory for stimulus properties in the primary visual cortex. PLoS Biol 7(12):1–19Google Scholar