Optical interconnection for neural networks by use of a self-imaging function

Area of Science:

• Optics
• Photonics
• Computational Neuroscience

Background:

• Optical interconnections are crucial for high-speed computing.
• Scaling self-transform functions offer potential for novel optical systems.
• Existing methods for optical neural networks can be complex.

Purpose of the Study:

• To develop a grating-lens unit capable of self-imaging on scale.
• To demonstrate the feasibility of using such units for optical interconnections in two-dimensional neural networks.

Main Methods:

• Development of a combined grating-lens unit.
• Arrangement of multiple units into an array for neural network interconnection.
• Experimental validation of the proposed system.

Main Results:

• The grating-lens unit successfully formed a scaling self-transform function.
• The array of units enabled effective optical interconnection for a two-dimensional neural network.
• The system proved to be feasible, simple, and easy to adjust optically.

Conclusions:

• The developed grating-lens unit is a viable component for optical scaling and self-imaging.
• This approach offers a simplified and practical solution for optical interconnections in neural network applications.