# US 2026/0073269 A1 · Optimization method and optimization device

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

- Number: US 2026/0073269 A1
- Document ID: US20260073269A1
- Status: Published application
- Kind: A1
- Application number: 19/108,288
- Filed: 2022-11-29
- Published: 2026-03-12
- Earliest claimed: 2022-11-29
- Assignees: TDK Corporation
- Applicants: TDK Corporation
- Inventors: Kaito ASAI
- Web page: https://patentgrep.com/patent/US20260073269A1

## CPC

- G06N 10/60: Quantum algorithms, e.g. based on quantum optimisation, quantum Fourier or Hadamard transforms
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- G06N 20/00: Machine learning
- G06N 99/00: Subject matter not provided for in other groups of this subclass
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

This optimization method has a first regression process, a first optimization process, and a first determination process. The first regression process regresses a first function using a first training data group formed from combinations of an explanatory variable column and an objective variable. The first optimization process performs optimization of the first function and obtains a first explanatory variable column that is an optimal solution and a first predicted value obtained by substituting the first explanatory variable column into the first function. The first determination process determines whether a relationship between the first predicted value and a threshold value satisfies a condition. In a case in which the condition is not satisfied, a combination of the first explanatory variable column and the first predicted value is added to the first training data group as one of the combinations of the explanatory variable column and the objective variable.

## Claims

### Claim 1

An optimization method comprising:
a first regression process of regressing a first function using a first training data group formed from combinations of an explanatory variable column and an objective variable;
a first optimization process of performing optimization of the first function and obtaining a first explanatory variable column that is an optimal solution of the first function and a first predicted value obtained by substituting the first explanatory variable column into the first function; and
a first determination process of determining whether a relationship between the first predicted value and a threshold value satisfies a condition,
wherein, in the first determination process, in a case in which the first predicted value does not satisfy the condition for the threshold value, a combination of the first explanatory variable column and the first predicted value is added to the first training data group as one of the combinations of the explanatory variable column and the objective variable.

### Claim 2 (depends on 1)

The optimization method according to claim 1, wherein, in the first determination process, until it is determined that the first predicted value satisfies the condition for the threshold value, a loop including the first regression process, the first optimization process, and the first determination process is repeated.

### Claim 3 (depends on 1)

The optimization method according to claim 1, wherein, in the first determination process, in a case in which the first predicted value satisfies the condition for the threshold value, it is determined whether the first explanatory variable column is executable on the basis of a restriction.

### Claim 4 (depends on 3)

The optimization method according to claim 3,
wherein, in a case in which it is determined that the first explanatory variable column is inexecutable on the basis of the restriction, the first predicted value is changed to a first correction value, and
wherein a combination of the first explanatory variable column and the first correction value is added to the first training data group as one of the combinations of the explanatory variable column and the objective variable, the combination of the first explanatory variable column and the first predicted value is removed from the first training data group, and the first training data group is set as a second training data group.

### Claim 5 (depends on 4)

The optimization method according to claim 4, wherein the first correction value is obtained using the following relationship equation (1).
f
′
⁢
(
s
)
=
(
1
-
r
)
×
f
⁢
(
s
)
+
r
×
t
(
1
)
In the relationship equation (1), f′(s) is the first correction value, f(s) is the first predicted value, r is an update rate satisfying 0<r<1, and t is the threshold value.

### Claim 6 (depends on 4)

The optimization method according to claim 4, further comprising:
a second regression process of regressing a second function using the second training data group;
a second optimization process of performing optimization of the second function and obtaining a second explanatory variable column that is an optimal solution of the second function and a second predicted value obtained by substituting the second explanatory variable column into the second function; and
a second determination process of determining whether a relationship between the second predicted value and a threshold value satisfies a condition,
wherein, in the second determination process, in a case in which the second predicted value does not satisfy the condition for the threshold value, a combination of the second explanatory variable column and the second predicted value is added to the second training data group as one of the combinations of the explanatory variable column and the objective variable, and
wherein, in the second determination process, in a case in which the second predicted value satisfies the condition for the threshold value, it is determined whether the second explanatory variable column is executable on the basis of the restriction.

### Claim 7 (depends on 6)

The optimization method according to claim 6, wherein, in the second determination process, until it is determined that the second predicted value satisfies the condition for the threshold value, a loop including the second regression process, the second optimization process, and the second determination process is repeated.

### Claim 8 (depends on 6)

The optimization method according to claim 6,
wherein, in a case in which it is determined that the second explanatory variable column is inexecutable on the basis of the restriction, the second predicted value is changed to a second correction value, and
wherein a combination of the second explanatory variable column and the second correction value is added to the second training data group as one of the combinations of the explanatory variable column and the objective variable, and the combination of the first explanatory variable column and the first correction value is removed from the second training data group.

### Claim 9 (depends on 6)

The optimization method according to claim 6, further comprising:
a third regression process of regressing a third function using a third training data group acquired by excluding the combination of the first explanatory variable column and the first correction value from the second training data group in a case in which it is determined that the second explanatory variable column is executable on the basis of the restriction;
a substitution process of obtaining a substitution value by substituting the second explanatory variable column into the third function; and
a third determination process of determining whether a relationship between the substation value and the threshold value satisfies a condition;
wherein, in a case in which the substitution value does not satisfy the condition for the threshold value, a combination of the second explanatory variable column and the substitution value is added to the first training data group as one of the combinations of the explanatory variable column and the objective variable.

### Claim 10 (depends on 1)

The optimization method according to claim 1, wherein the first regression process is performed by a factorization machine.

### Claim 11 (depends on 1)

The optimization method according to claim 1, wherein the first optimization process is performed using quantum annealing.

### Claim 12 (depends on 1)

The optimization method according to claim 1, wherein the first optimization process is performed using classical annealing.

### Claim 13

An optimization device comprising:
a storage area storing a training data group formed from combinations of an explanatory variable column and an objective variable;
a regression calculation area regressing a function by performing regression analysis of the training data group;
an optimization calculation area optimizing the function; and
a determination area determining whether or not the explanatory variable column and the predicted value obtained in the optimization calculation area satisfy a predetermined condition,
wherein, in a case in which the determination area determines that the explanatory variable column and the predicted value do not satisfy the predetermined condition, the explanatory variable column and the predicted value are added to the training data group of the storage area.
