Proposal of Carrier-Wave Reservoir Computing

Akira Hirose, Gouhei Tanaka, Seiji Takeda, Toshiyuki Yamane, Hidetoshi Numata, Naoki Kanazawa, Jean Benoit Héroux, Daiju Nakano, Ryosho Nakane

Abstract: Reservoir computing is highly compatible with physical waves such as optical wave and spin wave. In such wave-realized reservoir computing, signals are not limited to the raw signals, or baseband signals, but also expressed as frequency-shifted signals having a carrier. This paper proposes such a reservoir computing architecture, namely, carrier-wave reservoir computing. We present its construction and suitable learning dynamics with which we deal with the phase information explicitly. The merits of the proposed carrier-wave reservoir computing are its frequency-dependent processing functions, frequency-domain multiplexing ability and explicit phase-information utilization. It is also useful for material evaluation from the viewpoint of computational ability in the reservoir computing.