Adding filters to improve reservoir computer performance

Reservoir computers are a type of neuromorphic computer that may be built an analog system, potentially creating powerful small, light and consume little power. Typically reservoir is build by connecting together set nonlinear nodes into network; the difficult or expensive, however. This work shows how expanded adding functions to its output. The particular described here linear filters, but other possible. design construction filters well known, such easily implemented in hardware as field programmable gate arrays (FPGA's). effect on performance simulated for signal fitting problem, prediction problem classification problem.