Delay-embedding for recurrence in optical reservoir computing

Area of Science:

• Nonlinear dynamics
• Computational physics
• Machine learning

Background:

• Time-delay embedding enhances neural network predictions for nonlinear systems.
• Reservoir computing, a type of recurrent machine learning, benefits from time-delay embedding for improved performance and hyperparameter flexibility.

Purpose of the Study:

• To present an optical implementation of reservoir computing.
• To demonstrate the effectiveness of optical delay embedding for enhanced prediction performance in nonlinear systems.

Main Methods:

• An optical reservoir computing system was developed using a spatial light modulator.
• A recursive cascading effect was generated for optical delay embedding.
• The system was experimentally evaluated using the one-dimensional Kuramoto-Sivashinsky equation.

Main Results:

• The optical reservoir computing implementation achieved high parallelism and fast processing.
• Successful optical delay embedding was demonstrated through a recursive cascading effect.
• The experimental results showed superior performance compared to classical reservoir computing implementations.

Conclusions:

• Optical reservoir computing with spatial light modulators offers a powerful approach for nonlinear system prediction.
• The proposed optical delay embedding technique enhances prediction accuracy.
• This method provides a scalable and efficient alternative to traditional computational approaches.