Title:Reservoir computing for sensing: an experimental approach

Abstract:The increasing popularity of machine learning solutions puts increasing restrictions on this field if it is to penetrate more aspects of life. In particular, energy efficiency and speed of operation is crucial, inter alia in portable medical devices. The Reservoir Computing (RC) paradigm poses as a solution to these issues through foundation of its operation: the reservoir of states. Adequate separation of input information translated into the internal state of the reservoir, whose connections do not need to be trained, allow to simplify the readout layer thus significantly accelerating the operation of the system. In this brief review article, the theoretical basis of RC was first described, followed by a description of its individual variants, their development and state-of-the-art applications in chemical sensing and metrology: detection of impedance changes and ion sensing. Presented results indicate applicability of reservoir computing for sensing and validating the SWEET algorithm experimentally.