Figure 7
From: Scalable photonic reinforcement learning by time-division multiplexing of laser chaos
Diffusivity analysis of the time series to investigate the underlying structure. (a) Comparison of the ensemble averages of the time-averaged mean square displacements (ETMSDs), where the values for Chaos 1–4 agree with the superiority of the resulting decision-making performance. (b) Trajectories configured in a two-dimensional plane by 〈x(t)〉 and 〈x(t + D)〉 with D = 10,000 to examine the coverage by the time series. The trajectories spanned by the chaotic time series are uniformly distributed and are also quantitatively analysed using the smaller condition numbers shown in (a). (c) Distribution of the walker’s position at t = 100,000.
