Based on Daubechies (db) wavelet decomposition method, we propose to use the double photonic reservoir computers to enhance the chaotic synchronization quality of the two polarization components (X-PC and Y-PC) emitted by an optically pumped spin-VCSEL (the drive-VCSEL). The chaotic time series of two polarization components (PCs) from the drive-VCSEL, which serve as predictive targets, are decomposed into high-frequency and low-frequency components (multi-scale time series). This decomposition is performed using the first-type of Daubechies wavelet (db1-wavelet) decomposition. Our findings demonstrate that contrasted with a direct-prediction approach , the predictive errors in various intricate chaotic dynamics of the targets, outlined by permutation entropy, can be substantially minimized through db1-wavelet decomposition. As the layer number of the db1-wavelet decomposition increases, it noteworthily reduces the error in their predictive outcomes. Furthermore, under different parameter spaces, db1-wavelet decomposition considerably reduces the prediction errors for the targets and greatly enhances the quality of chaotic synchronization between pairs of X-PCs or Y-PCs, which are from the drive-VCSEL and reservoir-VCSEL. The further increase of the layer number of the db1-wavelet decomposition can trigger a more significant reduction in the prediction errors and improves the quality of chaotic synchronization. Interestingly, the use of different types of Daubechies wavelet decompositions does not substantially impact the prediction errors or the quality of chaotic synchronization. Our proposed model effectively addresses the issue of multi-scale optical time series prediction. The findings of this study can potentially be applied in enhancing the quality of optical chaotic synchronization in a photonic reservoir computing system.
Read full abstract