Reservoir computing based on transverse modes in a single optical waveguide.

Area of Science:

• Photonics and Optical Computing
• Reservoir Computing
• Waveguide Technology

Background:

• Conventional reservoir computing often faces limitations in scalability and node/connection density.
• Photonic implementations offer potential advantages in speed and parallelism.
• Existing waveguide-based systems can be constrained by planar fabrication limitations.

Purpose of the Study:

• To propose and validate a novel passive reservoir computing scheme using photonic waveguides.
• To investigate the role of optical modes and random defects as computational elements.
• To demonstrate the scalability of this approach with waveguide volume.

Main Methods:

• Development of a passive reservoir computing scheme based on transverse mode excitation and guiding in a large cross-section photonic waveguide.
• Utilizing optical modes as reservoir nodes and random defects as internode connections.
• Validation through numerical modeling and performance benchmarking on time-dependent image stream classification.

Main Results:

• Demonstrated that optical modes and defects effectively function as nodes and connections in a reservoir computing framework.
• Showcased the potential for significant scaling of nodes and connections with waveguide volume, surpassing planar designs.
• Successfully classified 2D analog/digital image streams, validating the scheme's practical applicability.

Conclusions:

• The proposed photonic waveguide reservoir computing scheme offers a scalable and effective alternative to conventional methods.
• Random defects play a crucial role in the performance and connectivity of the reservoir.
• This approach holds promise for advanced optical computing applications requiring high computational capacity.