Granted Patent B2

US 12,287,842 · App. 17/239,830 · Granted Apr 29, 2025

Inventors: Charles Roques-Carmes (Cambridge, MA); Yichen Shen (Cambridge, MA); Li Jing (Cambridge, MA); Tena Dubcek (Cambridge, MA); Scott A. Skirlo (Boston, MA); Hengameh Bagherianlemraski (Cambridge, MA); Marin Soljacic (Belmont, MA)

Assignee: Massachusetts Institute of Technology

G06F17/16G06N3/044G06N3/045G06N3/047G06N3/0675G06N3/084G06N7/01G06F17/18

Abstract

Claims (67)

1. A method of determining a ground state of an Ising model, the method comprising, at each of a plurality of time steps:

encoding a plurality of optical signals with amplitudes representing an initial spin state of the Ising model;

linearly transforming the plurality of optical signals to yield a plurality of linearly transformed optical signals;

perturbing the plurality of linearly transformed optical signals to yield a plurality of perturbed signals;

nonlinearly thresholding the plurality of perturbed signals to yield a plurality of nonlinearly thresholded signals; and

determining the ground state for the Ising model from the plurality of nonlinearly thresholded signals.

2. The method of claim 1 , wherein encoding the plurality of optical signals comprises:

splitting a pulse from a pulsed laser into N optical signals, where Nis a positive integer; and

modulating the amplitudes of the N optical signals to represent the initial spin state.

3. The method of claim 1 , wherein encoding the plurality of optical signals comprises synchronously transmitting the plurality of optical signals to a linear transformation unit.

4. The method of claim 1 , wherein linearly transforming the plurality of optical signals comprises performing a static linear transformation at each time step in the plurality of time steps.

5. The method of claim 1 , wherein the Ising model is represented by a Hamiltonian:

H

(

K

)

=

-

∑

1

≤

i

,

j

≤

N

K

ij

⁢

S

i

⁢

S

j

+

1

2

⁢

∑

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 linearly transforming the plurality of optical signals comprises performing a matrix multiplication with a matrix J based on the coupling matrix K.

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

7. The method of claim 1 , further comprising:

feeding the plurality of nonlinearly thresholded signals from a first time step in the plurality of time steps back into the plurality of optical signals in a second time step in the plurality of time steps.

8. The method of claim 1 , further comprising:

detecting a portion of each of the plurality of nonlinearly thresholded signals.

9. An optical processor comprising:

a substrate;

a first optical interference unit, integrated onto the substrate, to perform a first matrix multiplication on a first plurality of optical signals at a first time step and to perform a second matrix multiplication on a second plurality of optical signals at a second time step after the first time step;

a plurality of optical delay lines, integrated onto the substrate in photonic communication with the first optical interference unit, to produce a delayed copy of the first plurality of optical signals; and

a second optical interference unit, integrated onto the substrate in photonic communication with the plurality of optical delay lines, to perform a third matrix multiplication on the second plurality of optical signals and the delayed copy of the first plurality of optical signals.

10. The optical processor of claim 9 , further comprising:

a plurality of modulators, in photonic communication with the first optical interference unit, to modulate the first plurality of optical signals with an array of input values.

11. The optical processor of claim 10 , further comprising:

digital logic, operably coupled to the plurality of modulators, to parse digital values representing an image into a plurality of arrays of input values.

12. The optical processor of claim 11 , further comprising:

a plurality of photodetectors integrated on the substrate to detect an output of the optical processor.

13. The optical processor of claim 9 , wherein the first optical interference unit represents a first layer in a convolutional neural network and the second optical interference unit represents a second layer in the convolutional neural network.

Assignments (1)

Continuity (3)

References Cited (302)

Cited By (1)

US 12,524,661