Removing the masking procedure in delay-based reservoir computing

Delay-based reservoir computing (RC) offers a conceptually simple approach to construct an RC system, often requiring only few components. Within this framework, one sequentially injects input data into the RC system. This data is time-multiplexed with a preprocessing mask to implement virtual nodes within a delay line of the reservoir. However, this preprocessing procedure limits the computing bandwidth of the system. In this work, we introduce a novel photonic RC system based on a hybrid approach of two different architectures. We investigate a combination of a passive spatially distributed integrated photonic RC system and a temporally distributed RC system - containing a semiconductor laser - into a compound architecture. Based on numerical simulations, we show that we are able to remove the complex mask preprocessing procedure which can ultimately increase the computing bandwidth of delay-based reservoir computing.