Hybrid readout scheme for time delay reservoir computing using a semiconductor laser.

Area of Science:

• Nonlinear dynamics
• Computational intelligence
• Optoelectronics

Background:

• Reservoir computing (RC) leverages laser dynamics for efficient computation.
• Traditional RC methods often rely solely on optical intensity (OI) signals, requiring photodetection and potentially long feedback loops.
• This limits computational power and system complexity.

Purpose of the Study:

• To introduce and numerically investigate a hybrid readout reservoir computing (RC) system using a single semiconductor laser with optical feedback.
• To explore the integration of both optical intensity (OI) and terminal voltage (TV) signals for enhanced RC performance.
• To assess the system's effectiveness for time-series prediction tasks.

Main Methods:

• Numerical simulations of a semiconductor laser with optical feedback.
• Development of a hybrid readout RC system integrating OI and TV signals.
• Evaluation of reservoir performance using virtual nodes sampled from both OI and TV signals.

Main Results:

• Reservoirs using TV signals achieve performance comparable to OI-based methods, bypassing costly photodetection.
• Combining OI and TV signals effectively doubles the number of virtual nodes without changing the physical setup.
• The hybrid approach demonstrates comparable performance to single-signal methods but with potential for doubled processing speed.

Conclusions:

• The proposed hybrid readout RC system offers a cost-effective and efficient alternative to conventional methods.
• Integrating TV signals simplifies the RC system by eliminating the need for photoelectric conversion.
• This novel approach significantly enhances computational performance and processing speed in laser-based reservoir computing.