抄録
Recently, an information- processing method was proposed based on a semiconductor laser with time-delayed
optical feedback and optical injection. This method is called reservoir computing, which is a machine
learning paradigm based on information processing in the human brain. In this scheme, consistency
is a critical characteristic and represents the reproducibility of the responses of a dynamical system
when repeatedly driven by similar inputs. The convergence of consistent laser outputs is also important
for reservoir computing performance. In this study, we investigate the dependence of the convergence of
laser outputs on the initial optical frequency detuning between the two lasers. The convergence is quantitatively
evaluated using a conditional Lyapunov exponent. We also demonstrate reservoir computing
based on a semi-conductor laser and investigate the relationship between the performances of reservoir
computing and convergence of consistent laser output.
