Parallel information processing by a reservoir computing system based on a VCSEL subject to double optical feedback...

Area of Science:

• Nonlinear dynamics and optical systems
• Computational intelligence and machine learning
• Photonics and laser applications

Background:

• Reservoir computing (RC) offers a powerful framework for complex time series processing.
• Vertical-cavity surface-emitting lasers (VCSELs) exhibit rich nonlinear dynamics suitable for hardware implementations of RC.
• Parallel processing of distinct computational tasks is a key challenge in advanced computing.

Purpose of the Study:

• To propose and numerically investigate a novel reservoir computing scheme for simultaneous time series prediction and signal classification.
• To explore the impact of different optical feedback configurations on the parallel processing capabilities of the VCSEL-based RC system.
• To identify the optimal feedback parameters for maximizing the performance of the parallel RC tasks.

Main Methods:

• Utilizing a vertical-cavity surface-emitting laser (VCSEL) with double optical feedback and injection as the nonlinear node.
• Implementing parallel information processing based on the X and Y polarization components (X-PC, Y-PC) of the VCSEL.
• Numerically analyzing four feedback combination cases: polarization-preserved optical feedback (PP-OF) and polarization-rotated optical feedback (PR-OF).

Main Results:

• The parallel processing ability of the proposed RC system is significantly dependent on the chosen optical feedback frames.
• The best parallel processing performance for both prediction and classification tasks was achieved using polarization-preserved optical feedback (PP-OF) in both feedback loops.
• Optimized operation parameters yielded a low prediction error of 0.0289 and a signal classification error of 2.78 × 10^-5.

Conclusions:

• The VCSEL-based reservoir computing scheme effectively enables parallel processing of distinct computational tasks.
• Polarization-preserved optical feedback is crucial for optimizing the parallel performance of this specific RC system.
• This work demonstrates a promising approach for developing efficient, hardware-based parallel computing systems using photonic devices.