In this brief, we numerically demonstrate a photonic delay-based reservoir computing system, which processes, in parallel, two independent computational tasks even when the two tasks have unrelated input streams. Our approach is based on a single-longitudinal mode semiconductor ring laser (SRL) with optical feedback. The SRL emits in two directional optical modes. Each directional mode processes one individual task to mitigate possible crosstalk. We illustrate the feasibility of our scheme by analyzing the performance on two benchmark tasks: 1) chaotic time series prediction and 2) nonlinear channel equalization. We identify some feedback configurations for which the results for simultaneous prediction/classification indicate a good performance, but with slight degradation (as compared with the performance obtained for single task processing) due to nonlinear and linear interactions between the two directional modes of the laser. In these configurations, the system performs well on both tasks for a broad range of the parameters.

We design and fabricate a prototype scalable multichannel free-space interconnection module with the potential for Tb/s/spl middot/cm/sup 2/ aggregate bit-rate capacity over inter- and intra-MCM interconnection distances. The component is fabricated in a high quality optical plastic, PMMA, using deep proton lithography, an ion-based rapid prototyping technology. As a feasibility demonstration, data communication is achieved at 622 Mb/s per channel with a bit error rate smaller than 10/sup -13/ for 16 channels with an interchannel crosstalk lower than -22 dB. We perform a sensitivity analysis for misalignments and fabrication errors and study the fabrication issues of these components with injection molding techniques. Finally, we provide evidence that these modules can be mass fabricated with the required precision.