Optical Ising machines and optical convolutional neural networks

Patent metadata

Field	Value
Publication number	US-11017309-B2
Application number	US-201816032737-A
Country	US
Kind code	B2
Filing date	Jul 11, 2018
Priority date	Jul 11, 2017
Publication date	May 25, 2021
Grant date	May 25, 2021

Abstract

A photonic parallel network can be used to sample combinatorially hard distributions of Ising problems. The photonic parallel network, also called a photonic processor, finds the ground state of a general Ising problem and can probe critical behaviors of universality classes and their critical exponents. In addition to the attractive features of photonic networks—passivity, parallelization, high-speed and low-power—the photonic processor exploits dynamic noise that occurs during the detection process to find ground states more efficiently.

First claim

The invention claimed is:

1. A photonic processor for determining a spin distribution of a ground state of an Ising model, the photonic processor comprising:
a light source to generate a plurality of optical signals with amplitudes representing an initial spin state of the Ising model;
a matrix multiplication unit, in photonic communication with the light source, to linearly transform the plurality of optical signals into a plurality of linearly transformed optical signals;
a perturbation unit, operably coupled to the matrix multiplication unit, to perturb the plurality of linearly transformed optical signals to yield a plurality of perturbed signals; and
a nonlinear thresholding unit, operably coupled to the perturbation unit, to nonlinearly threshold the plurality of perturbed signals to yield a plurality of nonlinearly thresholded signals, the plurality of nonlinearly thresholded signals representing a spin distribution of the ground state of the Ising model.

2. The photonic processor of claim 1 , wherein the light source is configured to synchronously transmit the plurality of optical signals to the matrix multiplication unit.

3. The photonic processor of claim 1 , wherein the matrix multiplication unit is configured to perform a static linear transformation on the plurality of optical signals.

4. The photonic processor of claim 1 , wherein the Ising model is represented by a Hamiltonian:
H (K) =−Σ 1≤i,j≤N K ij S i S j +½Σ 1≤i,j≤N K ij ,

where K is a coupling matrix representing interactions between elements of the Ising model and S j represents a spin of a j th element of the Ising model, and wherein the linear transformation unit is configured to multiply the plurality of optical signals with a matrix J based on the coupling matrix K.

5. The photonic processor of claim 4 , wherein J is equal to the square root of {tilde over (K)}, which is a sum of K and a diagonal matrix Δ.

6. The photonic processor of claim 1 , further comprising:
a plurality of optical waveguides, in photonic communication with the matrix multiplication unit and the nonlinear thresholding unit, to feed the plurality of nonlinearly thresholded signals back into an input of the matrix multiplication unit.

7. The photonic processor of claim 1 , wherein the perturbation unit comprises at least one photodetector to detect and perturb the plurality of linearly transformed optical signals via measurement noise.

8. The photonic processor of claim 1 , wherein the light source comprises a pulsed laser having an output split into channels and an amplitude modulator to modulate each channel with a corresponding initial spin state.

9. The photonic processor of claim 1 , wherein the matrix multiplication unit comprises a set of interconnected Mach-Zehnder interferometers or ring resonators integrated onto a semiconductor substrate.

10. The photonic processor of claim 1 , wherein the nonlinear thresholding unit comprises a set of saturable absorbers.

11. The photonic processor of claim 1 , further comprising:
a dot product unit, operably coupled to the nonlinear threshold unit, to take a dot product of the plurality of thresholded signals.

12. The photonic processor of claim 1 , wherein the perturbation unit comprises at least one photodetector to convert the plurality of linearly transformed optical signals into electronic signals and is configured to perturb the electronic signals to yield the plurality of perturbed signals.

13. The photonic processor of claim 12 , wherein the nonlinear thresholding unit is configured to electronically threshold the plurality of perturbed signals.

Assignees

• Massachusetts Inst Technology

Inventors

• Roques-Carmes Charles
• Shen Yichen
• Jing Li
• Dubcek Tena
• Skirlo Scott A
• Bagherianlemraski Hengameh
• Soljacic Marin

Classifications

• G06N3/047

  Probabilistic or stochastic networks · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title