# US 2026/0065164 A1 · Transfer learning domain adaptation

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

- Number: US 2026/0065164 A1
- Document ID: US20260065164A1
- Status: Published application
- Kind: A1
- Application number: 19/315,271
- Filed: 2025-08-29
- Published: 2026-03-05
- Earliest claimed: 2024-09-05
- Assignees: Analog Devices, Inc.
- Applicants: Analog Devices, Inc.
- Inventors: Varun Kelkar; Melihcan Erol; Ravi Kiran Raman; Omer Tanovic
- Web page: https://patentgrep.com/patent/US20260065164A1

## CPC

- G06N 20/00: Machine learning
- G01R 31/2836: {Fault-finding or characterising (G01R31/2822 - G01R31/2831 take precedence)}
- G01R 31/367: Software therefor, e.g. for battery testing using modelling or look-up tables
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G01R: MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES (indicating correct tuning of resonant circuits H03J3/12)

## Abstract

System and techniques for adapting one domain to another to facilitate transfer learning in a machine learning (ML) model are described herein. Given data from a source domain and a target domain, moments can be mapped from the source domain to the target domain. Synthetic data that is constrained by the mapping can be created in the target domain. The synthetic data can be combined with the target domain data to create training data for a machine learning model to enable the model to be trained to accept target domain data as input and produce an output.

## Claims

### Claim 1

A device for transfer learning domain adaptation, the device comprising:
an interface configured to:
access first media that includes first data in a source domain; and
access second media that include second data in a target domain;
a memory including instructions; and
processing circuitry that, when in operation, is configured by the instructions to:
access, via the interface, the first data from the first media;
derive a moment from the first data, the moment being a quantitative measure of a statistical distribution;
access, via the interface, the second data from the second media;
create synthetic data for the target domain based on the moment;
create training data from the synthetic data and the second data; and
train a machine learning model using the training data, the machine learning model trained to accept target domain data as input and produce an output.

### Claim 2 (depends on 1)

The device of claim 1, wherein the second data is sparse in the target domain, the second data including a gap beyond a threshold.

### Claim 3 (depends on 2)

The device of claim 2, wherein the synthetic data fills a portion of the gap in the target domain to create the training data.

### Claim 4 (depends on 3)

The device of claim 3, wherein, to create the synthetic data based on the moment, the processing circuitry is configured to smear a portion of the second data to fill the portion of the gap constrained by the moment.

### Claim 5 (depends on 1)

The device of claim 1, wherein the moment is conditional.

### Claim 6 (depends on 1)

The device of claim 1, wherein, to create the training data from the synthetic data and the second data, the processing circuitry is configured to:
combine the synthetic data and the second data to create interim data;
create a transform to map the first data to the interim data via a training process to map a data point from the source domain to a corresponding data point in the target domain;
apply the transform to the first data to provide additional synthetic data; and
combine the additional synthetic data with the interim data.

### Claim 7 (depends on 6)

The device of claim 6, wherein the transform is a type of optimal transport from the source domain to the target domain.

### Claim 8 (depends on 1)

The device of claim 1, wherein the source domain is a first set of measurements for a first product and the target domain is a second set of measurements for a second product.

### Claim 9 (depends on 8)

The device of claim 8, wherein the first product and the second product are different products of a same type.

### Claim 10 (depends on 9)

The device of claim 9, wherein the same type is a battery, and wherein the first product has a different chemistry or a different form factor than the second product.

### Claim 11

A non-transitory machine readable medium including instructions for transfer learning domain adaptation, the instructions, when executed by processing circuitry, cause the processing circuitry to perform operations comprising:
accessing first media that includes first data in a source domain;
deriving a moment from the first data, the moment being a quantitative measure of a statistical distribution;
accessing second media that includes second data in a target domain;
creating synthetic data for the target domain based on the moment;
creating training data from the synthetic data and the second data; and
training a machine learning model using the training data, the machine learning model trained to accept target domain data as input and produce an output.

### Claim 12 (depends on 11)

The machine readable medium of claim 11, wherein the second data is sparse in the target domain, the second data including a gap beyond a threshold.

### Claim 13 (depends on 12)

The machine readable medium of claim 12, wherein the synthetic data fills a portion of the gap in the target domain to create the training data.

### Claim 14 (depends on 13)

The machine readable medium of claim 13, wherein creating the synthetic data based on the moment includes smearing a portion of the second data to fill the portion of the gap constrained by the moment.

### Claim 15 (depends on 11)

The machine readable medium of claim 11, wherein the moment is conditional.

### Claim 16 (depends on 11)

The machine readable medium of claim 11, wherein creating the training data from the synthetic data and the second data includes:
combining the synthetic data and the second data to create interim data;
creating a transform to map the first data to the interim data via a training process to map a data point from the source domain to a corresponding data point in the target domain;
applying the transform to the first data to provide additional synthetic data; and
combining the additional synthetic data with the interim data.

### Claim 17 (depends on 16)

The machine readable medium of claim 16, wherein the transform is a type of optimal transport from the source domain to the target domain.

### Claim 18 (depends on 11)

The machine readable medium of claim 11, wherein the source domain is a first set of measurements for a first product and the target domain is a second set of measurements for a second product.

### Claim 19 (depends on 18)

The machine readable medium of claim 18, wherein the first product and the second product are different products of a same type.

### Claim 20 (depends on 19)

The machine readable medium of claim 19, wherein the same type is a battery, and wherein the first product has a different chemistry or a different form factor than the second product.
