Optical computing device and computing method

Patent metadata

Field	Value
Publication number	US-12181909-B2
Application number	US-202217691945-A
Country	US
Kind code	B2
Filing date	Mar 10, 2022
Priority date	Sep 11, 2019
Publication date	Dec 31, 2024
Grant date	Dec 31, 2024

Abstract

An optical computing device and a computing method are provided for an optical Ising machine with high operation efficiency. The optical computing device includes a first spin array, an optical feedback network, and a second spin array, where the optical feedback network is separately connected to the first spin array and the second spin array. The first spin array may receive a first group of signals including N optical pulses or N electrical signals, and generate a first group of spin signals including N spin signals. The optical feedback network may receive the first group of spin signals, and generate, based on the first group of spin signals and specified first data, a first group of feedback signals including N feedback signals. The first spin array and the second spin array may process a plurality of signals in parallel, to improve computation efficiency of the optical computing device.

First claim

What is claimed is:

1. An optical computing device comprising:
a first spin array, configured to receive a first group of signals, and generate a first group of spin signals based on the first group of signals, wherein the first group of spin signals comprises N spin signals, and N is an integer not less than 2:
an optical feedback network, connected to the first spin array and configured to:
receive the first group of spin signals; and
generate a first group of feedback signals based on the first group of spin signals and specified first data, wherein the first group of feedback signals comprises N feedback signals; and
a second spin array, connected to the optical feedback network and configured to:
receive the first group of feedback signals and the first group of signals; and
generate a second group of spin signals based on the first group of feedback signals and the first group of signals, wherein the second group of spin signals comprises N spin signals.

2. The device according to claim 1 , wherein the optical feedback network is further configured to:
receive the second group of spin signals; and
generate a second group of feedback signals based on the second group of spin signals and the first data, wherein the second group of feedback signals comprises N feedback signals; and
the first spin array is further configured to:
receive the second group of feedback signals and a second group of signals, wherein a second group of optical pulses comprises N optical pulses, and an amplitude of the second group of optical pulses is greater than an amplitude of a first group of optical pulses; and
generate a third group of spin signals based on the second group of feedback signals and the second group of signals, wherein the third group of spin signals comprises N spin signals.

3. The device according to claim 1 , wherein the first spin array comprises N optical parametric oscillators, and each optical parametric oscillator in the first spin array is configured to receive one of the first group of signals.

4. The device according to claim 1 , wherein the second spin array comprises N optical parametric oscillators, each optical parametric oscillator in the second spin array is configured to receive one of the first group of signals and one of the first group of feedback signals, the N optical parametric oscillators in the second spin array are in a one-to-one correspondence with the N optical parametric oscillators in the first spin array, and signals received by optical parametric oscillators having a correspondence in the second spin array and the first spin array are the same.

5. The device according to claim 1 , wherein the optical feedback network comprises:
a first signal processor, configured to receive the first group of spin signals, and convert the first group of spin signals into a first optical signal matrix, wherein the first optical signal matrix comprises N×N optical pulses:
a problem loader, connected to the first signal processor and configured to generate a first feedback signal matrix based on the first optical signal matrix and the first data, wherein the first feedback signal matrix comprises N×N optical signals; and
a second signal processor, connected to the problem loader and configured to convert the first feedback signal matrix into the first group of feedback signals.

6. The device according to claim 5 , wherein
the second signal processor is further configured to receive the second group of spin signals, and convert the second group of spin signals into a second optical signal matrix, wherein the second optical signal matrix comprises N×N optical pulses:
the problem loader is further configured to generate a second feedback signal matrix based on the second optical signal matrix and the first data, wherein the second feedback signal matrix comprises N×N optical signals; and
the first signal processor is further configured to convert the second feedback signal matrix into the second group of feedback signals.

7. The device according to claim 1 , wherein the optical feedback network comprises a plurality of cascaded Mach-Zehnder (MZ) interferometer component, and each MZ interferometer component comprises a Mach-Zehnder interferometer and an optical switch disposed at an interval.

8. The device according to claim 1 , wherein the optical computing device further comprises:
a detector array, connected to the first spin array and configured to:
detect the third group of spin signals; and
in response to phases of the third group of spin signals being preset values, obtain a computation result of the first data based on the third group of spin signals.

9. The device according to claim 2 , wherein the first group of signals comprises the first group of optical pulses, the second group of signals comprises the second group of optical pulses, and the amplitude of the second group of optical pulses is greater than the amplitude of the first group of optical pulses.

10. A computing method, wherein the method is performed by an optical computing device: the optical computing device comprises a first spin array, a second spin array, and an optical feedback network separately connected to the first spin array and the second spin array; and the method comprises:
receiving, by the first spin array, a first group of signals, and generating a first group of spin signals based on the first group of signals, wherein the first group of spin signals comprises N spin signals, and N is an integer not less than 2;
generating, by the optical feedback network, a first group of feedback signals based on the first group of spin signals and specified first data, wherein the first group of feedback signals comprises N feedback signals; and
receiving, by the second spin array, the first group of signals and the first group of feedback signals, and generating a second group of spin signals based on the first group of feedback signals and the first group of signals, wherein the second group of spin signals comprises N spin signals.

11. The method according to claim 10 , further comprising:
receiving, by the optical feedback network, the second group of spin signals sent by the second spin array;
generating, by the optical feedback network, a second group of feedback signals based on the second group of spin signals and the first data, wherein the second group of feedback signals comprises N feedback signals;
receiving, by the first spin array, a second group of signals; and
generating, by the first spin array, a third group of spin signals based on the second group of feedback signals and the second group of signals, wherein the third group of spin signals comprises N spin signals.

12. The method according to claim 11 , wherein the optical computing device further comprises a detector array, and the method further comprises:
detecting, by the detector array, the third group of spin signals; and
in response to phases of the third group of spin signals being preset values, obtaining a computation result of the first data based on the third group of spin signals.

13. The method according to claim 10 , wherein the receiving; the first group of signals comprises:
receiving, by each optical parametric oscillator in the first spin array, one of the first group of signals, wherein the first spin array comprises N optical parametric oscillators.

14. The method according to claim 13 , wherein the second spin array comprises N optical parametric oscillators; and the receiving the first group of signals and the first group of feedback signals comprises:
receiving, by each optical parametric oscillator in the second spi

Assignees

• Huawei Tech Co Ltd

Inventors

• Yun Zhiqiang
• Zhang Xiang
• Dong Xiaowen

Classifications

• G06N3/044

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

• G06N3/047

  Probabilistic or stochastic networks · CPC title

• G06F15/7817 Primary

  Specially adapted for signal processing, e.g. Harvard architectures · CPC title

• G06E3/008 Primary

  Matrix or vector computation · CPC title

• G06N3/0675 Primary

  using electro-optical, acousto-optical or opto-electronic means · CPC title