Systems and methods for quantum processing of data

Patent metadata

Field	Value
Publication number	US-9727824-B2
Application number	US-201414316372-A
Country	US
Kind code	B2
Filing date	Jun 26, 2014
Priority date	Jun 28, 2013
Publication date	Aug 8, 2017
Grant date	Aug 8, 2017

Abstract

Systems, methods and aspects, and embodiments thereof relate to unsupervised or semi-supervised features learning using a quantum processor. To achieve unsupervised or semi-supervised features learning, the quantum processor is programmed to achieve Hierarchal Deep Learning (referred to as HDL) over one or more data sets. Systems and methods search for, parse, and detect maximally repeating patterns in one or more data sets or across data or data sets. Embodiments and aspects regard using sparse coding to detect maximally repeating patterns in or across data. Examples of sparse coding include L0 and L1 sparse coding. Some implementations may involve appending, incorporating or attaching labels to dictionary elements, or constituent elements of one or more dictionaries. There may be a logical association between label and the element labeled such that the process of unsupervised or semi-supervised feature learning spans both the elements and the incorporated, attached or appended label.

First claim

What is claimed is:

1. A method of using a quantum processor to identify maximally repeating patterns in data via Hierarchical Deep Learning (HDL), the method comprising:
receiving a data set of data elements at a non-quantum processor;
preprocessing the data set of data elements to generate a preprocessed data set;
formulating an objective function based on the preprocessed data set via the non-quantum processor, wherein the objective function includes a loss term to minimize difference between a first representation of the preprocessed data set and a second representation of the preprocessed data set, and includes a regularization term to minimize any complications in the objective function;
casting a first set of weights in the objective function as variables using the non-quantum processor;
setting a first set of values for a dictionary of the objective function using the non-quantum processor, wherein the first set of values for the dictionary is constrained such that the objective function matches a connectivity structure of the quantum processor; and
interacting with the quantum processor, via the non-quantum processor, to minimize the objective function.

2. The method of claim 1 wherein the data set of data elements comprises an image data set of image data elements, and preprocessing the data set of data elements comprises normalizing at least one of a contrast or a brightness of the image data elements of the image data set.

3. The method of claim 2 wherein preprocessing the data set of data elements comprises whitening the normalized image data elements of the image data set.

4. The method of claim 3 wherein whitening the normalized image data elements of the image data set comprises applying zero phase component analysis (ZCA) whitening to the normalized data elements of the image data set.

5. The method of claim 2 wherein preprocessing the data set of data elements comprises reducing a dimensionality of the normalized image data elements of the image data set.

6. The method of claim 5 wherein reducing a dimensionality of the normalized image data elements of the image data set comprises applying principal component analysis (PCA) to the normalized data elements of the image data set.

7. The method of claim 2 wherein preprocessing the data set of data elements comprises reducing a dimensionality of the normalized image data elements of the image data set and whitening the normalized image data elements of the image data set.

8. The method of claim 1 wherein the data set of data elements comprises an image data set of image data elements, the method comprising:
segmenting each of the image data elements into one or more disjoint regions.

9. The method of claim 8 , further comprising:
receiving, by the non-quantum processor, a segmentation parameter indicative of a segmentation characteristic, wherein the segmenting each of the image data elements into one or more disjoint regions is at least partially based on the received segmentation parameter.

10. The method of claim 1 wherein formulating an objective function includes formulating the objective function where the regularization term is governed by an L0-norm form.

11. The method of claim 1 wherein formulating an objective function includes formulating the objective function where the regularization term is governed by an L1-norm form.

12. The method of claim 1 wherein the regularization term includes a regularization parameter, and formulating an objective function comprises selecting a value for the regularization parameter to control a sparsity of the objective function.

13. The method of claim 1 wherein receiving a data set of data elements at a non-quantum processor comprises receiving image data and audio data.

14. The method of claim 1 wherein interacting with the quantum processor, via the non-quantum processor, to minimize the objective function comprises:
optimizing the objective function for the first set of values for the weights in the objective function based on the first set of values for the dictionary.

15. The method of claim 14 wherein optimizing the objective function for a first set of values for the weights includes mapping the objective function to a first quadratic unconstrained binary optimization (“QUBO”) problem and using the quantum processor to at least approximately minimize the first QUBO problem, wherein using the quantum processor to at least approximately minimize the first QUBO problem includes using the quantum processor to perform at least one of adiabatic quantum computation or quantum annealing.

16. The method of claim 14 wherein interacting with the quantum processor, via the non-quantum processor, to minimize the objective function further comprises optimizing the objective function for a second set of values for the weights based on a second set of values for the dictionary, wherein optimizing the objective function for a second set of values for the weights includes mapping the objective function to a second QUBO problem and using the quantum processor to at least approximately minimize the second QUBO problem.

17. The method of claim 14 wherein interacting with the quantum processor, via the non-quantum processor, to minimize the objective function further comprises optimizing the objective function for a second set of values for the dictionary based on the first set of values for the weights, wherein optimizing the objective function for a second set of values for the dictionary includes using the non-quantum processor to update at least some of the values for the dictionary.

18. The method of claim 17 wherein interacting with the quantum processor, via the non-quantum processor, to minimize the objective function further comprises optimizing the objective function for a third set of values for the dictionary based on the second set of values for the weights, wherein optimizing the objective function for a third set of values for the dictionary includes using the non-quantum processor to update at least some of the values for the dictionary.

19. The method of claim 18 , further comprising:
optimizing the objective function for a t th set of values for the weights, where t is an integer greater than 2, based on the third set of values for the dictionary, wherein optimizing the objective function for a t th set of values for the weights includes mapping the objective function to a t th QUBO problem and using the quantum processor to at least approximately minimize the t th QUBO problem; and
optimizing the objective function for a (t+1) th set of values for the dictionary based on the t th set of values for the weights, wherein optimizing the objective function for a (t+1) th set of values for the dictionary includes using the non-quantum processor to update at least some of the values for the dictionary.

20. The method of claim 19 , further comprising optimizing the objective function for a (t+1) th set of values for the weights based on the (t+1) th set of values for the dictionary, wherein optimizing the objective function for a (t+1) th set of values for the weights includes mapping the objective function to a (t+1) QUBO problem and using the quantum processor to at least approximately minimize the (t+1) th QUBO problem.

21. The method of claim 19 wherein optimizing the objective function for a (t+1) th set of values for the dictionary based on the t th set of values for the weights and optimizing the objective function for a (t+1) th set of values for the we

Assignees

• D Wave Systems Inc

Inventors

• Rose Geordie
• Gildert Suzanne
• Macready William G
• Walliman Dominic Christoph

Classifications

• G06V10/772

  Determining representative reference patterns, e.g. averaging or distorting patterns; Generating dictionaries · CPC title

• G06V10/7715

  Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods · CPC title

• G06V10/955

  using specific electronic processors · CPC title

• G06F18/2135

  based on approximation criteria, e.g. principal component analysis · CPC title

• G06F18/2136

  based on sparsity criteria, e.g. with an overcomplete basis · CPC title