# US 2022/0024474 A1 · Computer-implemented method for machine learning for operating a vehicle component, and method for operating a vehicle component

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

- Number: US 2022/0024474 A1
- Document ID: US20220024474A1
- Status: Published application
- Kind: A1
- Application number: 17/380,421
- Filed: 2021-07-20
- Published: 2022-01-27
- Earliest claimed: 2020-07-21
- Assignees: Volkswagen Aktiengesellschaft
- Applicants: Volkswagen Aktiengesellschaft
- Inventors: Jonas Kaste; Kristof van Ende
- Web page: https://patentgrep.com/patent/US20220024474A1

## CPC

- B60W 50/06: Improving the dynamic response of the control system, e.g. improving the speed of regulation or avoiding hunting or overshoot
- G06N 3/08: Learning methods
- G06N 3/0499: Feedforward networks
- G06N 3/09: Supervised learning
- G06N 20/00: Machine learning
- B60W: CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

The invention relates to a method for preparing or generating a training data set for machine learning to operate a vehicle component. Provided multidimensional data points are divided up in a first step by dividing up the plurality of data points into multidimensional clusters by using a cluster algorithm. Then a training data set is generated by selecting data points from the basic training data set. The selection comprises determining a smallest cluster among the plurality of clusters with the lowest number of data points. Furthermore, at least one subset of the data points of the smallest cluster is provided for the training data set. In another step, a subset of data points is selected from each of the other clusters for the training data set, wherein the number of selected data points of each other cluster corresponds to the number of selected data points of the smallest cluster.

## Claims

### Claim 1

A computer-implemented method for machine learning to operate a vehicle component comprising:
providing a plurality of multidimensional data points;
generating a basic training data set by dividing up the plurality of data points into multidimensional clusters using a cluster algorithm such that each data point is assigned precisely to one cluster;
generating a training data set by selecting data points from the basic training data set, comprising:
determining a smallest cluster among the plurality of clusters with the lowest number of data points;
selecting at least one subset of the data points of the smallest cluster for the training data set; and
selecting a subset of data points from each of the other clusters for the training data set such that the number of selected data points of each other cluster corresponds to the number of selected data points of the smallest cluster.

### Claim 2 (depends on 1)

The method of claim 1, furthermore comprising: saving the unselected data points of the smallest cluster and/or the other clusters in a data memory as test and/or validation data for an algorithm for machine learning.

### Claim 3 (depends on 1)

The method of claim 1, furthermore comprising the selection of the totality of all data points of the smallest cluster for the training data set.

### Claim 4 (depends on 1)

The method of claim 1, wherein the selection of data points from the smallest cluster and/or the other clusters is random.

### Claim 5 (depends on 1)

The method of claim 1, wherein the data points are selected from the smallest cluster and/or the other clusters such that the selected data points of a cluster have a maximum distance from each other.

### Claim 6 (depends on 1)

The method of claim 1, furthermore comprising a training of an algorithm for machine learning using the generated training data set.

### Claim 7 (depends on 6)

The method of claim 6, furthermore comprising the determination of the performance of the individual clusters by means of a quality criterion by using validation data.

### Claim 8 (depends on 7)

The method of claim 7, furthermore comprising the determination of the performance of the individual clusters by the data points from the group of unselected data points of the associated clusters.

### Claim 9 (depends on 7)

The method of claim 7, wherein the quality criterion comprises a mean square deviation above a threshold value of an output value relative to the target value.

### Claim 10 (depends on 7)

The method of claim 7, furthermore comprising the creation of a modified training data set by adding data points to a cluster for which the quality criterion has not been met.

### Claim 11 (depends on 10)

The method of claim 10, wherein the added data points are taken from the group of unselected data points of the cluster for which the quality criterion is not met.

### Claim 12 (depends on 10)

The method of claim 10, furthermore comprising the repetition of the modification of the training data set until the quality criterion is satisfied for all clusters.

### Claim 13 (depends on 1)

A method for operating a vehicle component comprising the operation of the vehicle component based on an output of an algorithm for machine learning trained according to the method of claim 1.

### Claim 14 (depends on 13)

A control unit for a vehicle, configured to execute the method of claim 13.

### Claim 15 (depends on 14)

A vehicle comprising the control unit of claim 14.

### Claim 16 (depends on 2)

The method of claim 2, furthermore comprising the selection of the totality of all data points of the smallest cluster for the training data set.

### Claim 17 (depends on 2)

The method of claim 2, wherein the selection of data points from the smallest cluster and/or the other clusters is random.

### Claim 18 (depends on 3)

The method of claim 3, wherein the selection of data points from the smallest cluster and/or the other clusters is random.

### Claim 19 (depends on 2)

The method of claim 2, wherein the data points are selected from the smallest cluster and/or the other clusters such that the selected data points of a cluster have a maximum distance from each other.

### Claim 20 (depends on 1)

A control unit for a vehicle, configured to execute the method of claim 1.
