Data encoding and classification

Patent metadata

Field	Value
Publication number	US-10977558-B2
Application number	US-201916265375-A
Country	US
Kind code	B2
Filing date	Feb 1, 2019
Priority date	Feb 1, 2018
Publication date	Apr 13, 2021
Grant date	Apr 13, 2021

Abstract

In a method and apparatus for training a computer system for use in classification of an image by processing image data representing the image, image data are compressed and then loaded into a programmable quantum annealing device that includes a Restricted Boltzmann Machine. The Restricted Boltzmann Machine is trained to act as a classifier of image data, thereby providing a trained Restricted Boltzmann Machine; and, the trained Restricted Boltzmann Machine is used to initialize a neural network for image classification thereby providing a trained computer system for use in classification of an image.

First claim

The invention claimed is:

1. A method of training a computer system for use in classification of an image by processing image data representing the image, comprising:
compressing the image data by training an artificial neural network using the image data in an unsupervised manner, wherein said training said artificial neural network comprises training, an auto-encoder, and said training an auto-encoder comprises providing an auto-encoder comprising an encoder part having an input layer, and a decoder part having an output layer, and one or more hidden lavers connecting the encoder art to the decoder part, wherein the output layer has the same number of nodes as the input layer;
loading the compressed image data onto a programmable quantum annealing device comprising a Restricted Boltzmann Machine;
training the Restricted Boltzmann Machine to act as a classifier of image data, thereby providing a trained Restricted Boltzmann Machine; and,
using the trained Restricted Boltzmann Machine to initialize a neural network for image classification thereby providing a trained computer system for use in classification of an image.

2. The method according to claim 1 , further comprising using the encoder part to generate a feature vector which is a compressed representation of the image data.

3. The method according to claim 2 , wherein said loading the compressed image data onto the programmable quantum annealing device comprises loading said feature vector onto the programmable quantum annealing device.

4. A computer system or learning machine comprising a trained computer system which is trained according to claim 1 .

5. A computer system for classifying imaging data by processing image data representing the image, comprising:
a programmable quantum annealing device comprising a Restricted Boltzmann Machine;
a neural network for image classification;
a processor configured to:
compress the image data and to load the compressed image data onto said programmable quantum annealing device;
train the Restricted Boltzmann Machine to act as a classifier of image data, thereby providing a trained Restricted Boltzmann Machine; and
use the trained Restricted Boltzmann Machine to initialize the neural network thereby training the computer system for use in classification of an image; and

an artificial neural network comprising an auto-encoder and arranged for compressing said image data in an unsupervised manner, wherein said auto-encoder comprises an encoder part comprising an input layer, and a decoder part comprising an output layer, and one or more hidden layers connecting the encoder part to the decoder part, wherein the output layer has the same number of nodes as the input layer.

6. A computer system according to claim 5 , wherein the encoder part is arranged to generate a feature vector which is a compressed representation of the image data.

7. A non-transitory, computer-readable data storage medium encoded with programming instructions, said storage medium being loaded into a computer system and said programming instructions causing said computer system to classify an image by processing image data representing the image, with said programming instructions causing said computer system to:
compress the image data in an unsupervised manner using an artificial neural network comprising an auto-encoder, wherein said auto-encoder comprises an encoder part comprising an input layer, and a decoder part comprising an output layer, and one or more hidden layers connecting the encoder part to the decoder part, wherein the output layer has the same number of nodes as the input layer;
load the compressed image data onto a programmable quantum annealing device comprising a Restricted Boltzmann Machine;
train the Restricted Boltzmann Machine to act as a classifier of image data, thereby providing a trained Restricted Boltzmann Machine; and
use the trained Restricted Boltzmann Machine to initialize a neural network for image classification thereby provide a trained computer system for use in classification of an image.

Assignees

• Siemens Healthcare Ltd
• Siemens Healthcare Gmbh

Inventors

• Herbster Mark
• Mountney Peter
• Piat Sebastien
• Severini Simone

Classifications

• G06N3/047

  Probabilistic or stochastic networks · CPC title

• G06N3/0495

  Quantised networks; Sparse networks; Compressed networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title