18249726. SCALE-, SHIFT-, AND ROTATION-INVARIANT DIFFRACTIVE OPTICAL NETWORKS simplified abstract (THE REGENTS OF THE UNIVERSITY OF CALIFORNIA)

From WikiPatents
Jump to navigation Jump to search

SCALE-, SHIFT-, AND ROTATION-INVARIANT DIFFRACTIVE OPTICAL NETWORKS

Organization Name

THE REGENTS OF THE UNIVERSITY OF CALIFORNIA

Inventor(s)

Aydogan Ozcan of Los Angeles CA (US)

Deniz Mengu of Los Angeles CA (US)

Yair Rivenson of Los Angeles CA (US)

SCALE-, SHIFT-, AND ROTATION-INVARIANT DIFFRACTIVE OPTICAL NETWORKS - A simplified explanation of the abstract

This abstract first appeared for US patent application 18249726 titled 'SCALE-, SHIFT-, AND ROTATION-INVARIANT DIFFRACTIVE OPTICAL NETWORKS

Simplified Explanation

The abstract describes a method for creating an optical neural network that can process input object images or optical signals while being invariant to object transformations. The method involves training a software-based neural network model to perform specific optical functions for a multi-layer optical network. The training includes using input images or signals with random transformations and adjusting transmission/reflection coefficients for each layer until optimized coefficients are obtained. A physical embodiment of the network is then created with substrate layers that match the optimized coefficients.

  • The method involves training a software-based neural network model to perform optical functions for an optical neural network.
  • The training includes using input images or signals with random transformations.
  • Transmission/reflection coefficients for each layer of the network are adjusted iteratively until optimized coefficients are obtained.
  • A physical embodiment of the network is created with substrate layers matching the optimized coefficients.

Potential Applications

  • Image recognition and processing
  • Optical signal processing
  • Pattern recognition
  • Machine learning

Problems Solved

  • Invariance to object transformations
  • Efficient processing of input object images or optical signals
  • Optimization of transmission/reflection coefficients for each layer

Benefits

  • Improved accuracy and efficiency in processing object images or optical signals
  • Ability to handle object transformations without affecting performance
  • Physical embodiment of the network allows for practical implementation


Original Abstract Submitted

A method of forming an optical neural network for processing an input object image or optical signal that is invariant to object transformations includes training a software-based neural network model to perform one or more specific optical functions for a multi-layer optical network having physical features located in each of the layers of the optical neural network. The training includes feeding different input object images or optical signals that have random transformations or shifts and computing at least one optical output of optical transmission and/or reflection through the optical neural network using an optical wave propagation model and iteratively adjusting transmission/reflection coefficients for each layer until optimized transmission/reflection coefficients are obtained. A physical embodiment of the optical neural network is then made that has a plurality of substrate layers having physical features that match the optimized transmission/reflection coefficients obtained by the trained neural network model.

Retrieved from "http://wikipatents.org/index.php?title=18249726._SCALE-,_SHIFT-,_AND_ROTATION-INVARIANT_DIFFRACTIVE_OPTICAL_NETWORKS_simplified_abstract_(THE_REGENTS_OF_THE_UNIVERSITY_OF_CALIFORNIA)&oldid=18706"