Photonic time-delayed reservoir computing based on series-coupled microring resonators with high memory capacity

Area of Science:

• Photonics
• Computer Science
• Materials Science

Background:

• On-chip microring resonators (MRRs) are explored for time-delayed reservoir computing (RC).
• Single MRRs lack sufficient memory capacity for complex computations.
• Existing RC systems with optical feedback require impractically long waveguides.

Purpose of the Study:

• To propose a novel time-delayed reservoir computing system using silicon-based nonlinear and linear MRRs.
• To enhance memory capacity and computational performance in photonic RC.
• To reduce the physical footprint of high-performance RC systems.

Main Methods:

• Utilizing a silicon-based nonlinear microring resonator (MRR) coupled with an array of high-quality factor linear MRRs.
• Implementing a time-delayed reservoir computing architecture.
• Quantitatively assessing performance on Narma 10, Mackey-Glass, and Santa Fe chaotic time-series prediction tasks.

Main Results:

• The proposed RC system demonstrates comparable performance to optical feedback MRR systems for tasks with high memory demands (e.g., Narma 10).
• The new system's dimension is at least 350 times smaller than traditional optical feedback systems.
• High-quality factor linear MRRs effectively provide the necessary memory capacity.

Conclusions:

• The hybrid MRR approach offers a scalable and efficient solution for photonic reservoir computing.
• This design significantly reduces the complexity and size of RC systems.
• The study provides a foundation for integrating advanced photonic RC for diverse computational challenges.