Performance optimization research of reservoir computing system based on an optical feedback semiconductor laser...

Area of Science:

• Optoelectronics
• Nonlinear Dynamics
• Computational Science

Background:

• Reservoir computing (RC) offers a powerful framework for complex system modeling.
• Semiconductor lasers (SLs) with optical feedback are explored as a physical basis for RC systems.
• Electrical information injection is a key method for interfacing with SL-based RC.

Purpose of the Study:

• To numerically investigate the performance of an SL-based RC system with electrical information injection.
• To identify optimal operating parameters for the SL feedback delay time and strength.
• To evaluate the system's effectiveness in time series prediction and nonlinear channel equalization.

Main Methods:

• Numerical simulations of an optical feedback SL system.
• Implementation of Santa Fe time series prediction benchmark.
• Testing of nonlinear channel equalization with varying signal-to-noise ratios (SNRs).
• Optimization of feedback delay time and strength based on SL current.

Main Results:

• Feedback delay time and strength critically impact RC system performance.
• Optimized parameters enable good performance for SL biased between 1.1-3.5 times threshold.
• Prediction errors below 0.01 achieved for Santa Fe tests.
• Symbol error rates (SERs) on the order of 10^-5 for channel equalization at 32 dB SNR.
• Increasing SL current enhances information processing rate for a given RC performance.

Conclusions:

• Optimized optical feedback in SLs provides a robust platform for reservoir computing.
• The system demonstrates high accuracy in time series prediction and channel equalization.
• Adjusting SL current offers a method to increase the information processing speed of the RC system.