# US 12,223,433 B2 · Unsupervised learning method for general inverse problem and apparatus therefor

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,223,433 B2
- Document ID: US12223433B2
- Status: Granted
- Kind: B2
- Application number: 17/332,696
- Filed: 2021-05-27
- Published: 2022-01-27
- Granted: 2025-02-11
- Earliest claimed: 2020-07-20
- Assignees: Korea Advanced Institute of Science and Technology
- Applicants: Korea Advanced Institute of Science and Technology
- Inventors: JongChul Ye; Byeongsu Sim; Gyutaek Oh
- Web page: https://patentgrep.com/patent/US12223433B2

## CPC

- G06N 3/088: Non-supervised learning, e.g. competitive learning
- G06F 18/214: Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F 18/24133: {Distances to prototypes}
- G06N 3/045: Combinations of networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/047: Probabilistic or stochastic networks
- G06N 3/0475: Generative networks
- G06N 3/048: Activation functions
- G06N 3/094: Adversarial learning
- G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

Disclosed are an unsupervised learning method and an apparatus therefor applicable to general inverse problems. An unsupervised learning method applicable to inverse problems includes receiving a training data set and training an unsupervised learning-based neural network generated based on an optimal transport theory and a penalized least square (PLS) approach using the training data set, wherein the receiving of the training data set includes receiving the training data set including unmatched data.

## Claims

### Claim 1

An unsupervised learning method applicable to inverse problems, comprising:
receiving a training data set; and
training an unsupervised learning-based neural network generated based on an optimal transport theory and a penalized least square (PLS) approach using the training data set, wherein the neural network includes:
a first neural network configured to convert a first magnetic resonance image (MRI) obtained, as an input, from an intermittent Fourier spatial coefficient into a second MRI corresponding to a complete Fourier spatial coefficient;
a Fourier transform unit configured to output a third MRI corresponding to the first MRI by applying a Fourier transform and an inverse Fourier transform to the second MRI; and
a second neural network configured to discriminate between the second MRI and an actual MRI for the second MRI.

### Claim 2 (depends on 1)

The unsupervised learning method of claim 1, wherein the receiving of the training data set includes receiving the training data set including unmatched data.

### Claim 3 (depends on 1)

The unsupervised learning method of claim 1, wherein the neural network is trained via unsupervised learning based on a cyclic loss between the first image and the third image and an adversarial loss between the second image and the actual image.

### Claim 4 (depends on 1)

The unsupervised learning method of claim 1, wherein the neural network includes any one of a neural network based on a convolution framelet and a neural network including a pooling layer and an unpooling layer.

### Claim 5

An image processing method, comprising:
receiving a first image; and
reconstructing the first image as a second image corresponding to the first image using an unsupervised learning-based neural network generated based on an optimal transport theory and a penalized least square (PLS) approach, wherein the neural network includes:
a first neural network configured to convert a first magnetic resonance image (MRI) obtained, as an input, from an intermittent Fourier spatial coefficient into a second MRI corresponding to a complete Fourier spatial coefficient;
a Fourier transform unit configured to output a third MRI corresponding to the first MRI by applying a Fourier transform and an inverse Fourier transform to the second MRI; and
a second neural network configured to discriminate between the second MRI and an actual MRI for the second MRI.

### Claim 6

An unsupervised learning apparatus applicable to inverse problems, comprising:
a receiving unit configured to receive a training data set; and
a training unit configured to train an unsupervised learning-based neural network generated based on an optimal transport theory and a penalized least square (PLS) approach using the training data set, wherein the neural network includes:
a first neural network configured to convert a first magnetic resonance image (MRI) obtained, as an input, from an intermittent Fourier spatial coefficient into a second MRI corresponding to a complete Fourier spatial coefficient;
a Fourier transform unit configured to output a third MRI corresponding to the first MRI by applying a Fourier transform and an inverse Fourier transform to the second MRI; and
a second neural network configured to discriminate between the second MRI and an actual MRI for the second MRI.

### Claim 7 (depends on 6)

The unsupervised learning apparatus of claim 6, wherein the receiving unit receives the training data set including unmatched data.

### Claim 8 (depends on 6)

The unsupervised learning apparatus of claim 6, wherein the neural network is trained via unsupervised learning based on a cyclic loss between the first image and the third image and an adversarial loss between the second image and the actual image.

### Claim 9 (depends on 6)

The unsupervised learning apparatus of claim 6, wherein the neural network includes any one of a neural network based on a convolution framelet and a neural network including a pooling layer and an unpooling layer.