# US 2026/0289425 A1 · Jointly predicting multiple individual-level features from aggregate data

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

- Number: US 2026/0289425 A1
- Document ID: US20260289425A1
- Status: Published application
- Kind: A1
- Application number: 19/678,532
- Filed: 2026-05-15
- Published: 2026-09-24
- Earliest claimed: 2022-02-25
- Assignees: Adobe Inc.
- Applicants: Adobe Inc.
- Inventors: Gaurav SINHA; Rashul Chutani; Shaurya Goel; Isha Chaudhary; Simarpreet Singh Saluja; Vibhor Porwal
- Web page: https://patentgrep.com/patent/US20260289425A1

## CPC

- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- G06F 18/213: Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F 18/214: Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F 18/24: Classification techniques
- G06Q 30/0246: {Traffic}
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR

## Abstract

An analytics system jointly predicts values for multiple unobserved individual-level features using aggregate data for those features. Given a dataset, a transformation is applied to individual-level information for the dataset to generate transformed data in a higher dimensional space. Bag-wise mean embeddings are generated using the transformed data. The bag-wise mean embeddings and aggregate data for unobserved individual-level features for the dataset are used to train a model to jointly predict values for the unobserved individual-features for data instances. In particular, a given data instance can be transformed to a representation in a higher dimensional space. Given this representation, the trained model predicts values for the unobserved individual-level features for the data instance, and the data instance can be augmented with the predicted values.

## Claims

### Claim 1

One or more computer storage media storing computer-useable instructions that, when used by a computing device, cause the computing device to perform operations, the operations comprising:
accessing a dataset stored on a computer storage device, the dataset comprising a plurality of observed individual-level features;
accessing aggregate data associated with the dataset, the aggregate data corresponding to two or more unobserved individual-level features and aggregated over groups defined by a grouping feature of the dataset;
generating, based on the observed individual-level features, a transformed representation of at least a portion of the dataset, the transformed representation encoding relationships among data instances within each group defined by the grouping feature;
training, using the transformed representation and the aggregate data, a joint predictive model configured to learn inter-feature dependencies among the two or more unobserved individual-level features;
applying the trained joint predictive model to input corresponding to a first individual data instance to generate predicted values for each of the two or more unobserved individual-level features; and
augmenting the dataset by associating the predicted values with the first individual data instance.

### Claim 2 (depends on 1)

The one or more computer storage media of claim 1, wherein generating the transformed representation of at least the portion of the dataset comprises:
transforming values of the observed individual-level features of each data instance of the dataset to generate transformed data in a higher dimensional space for each data instance; and
for each distinct value of a first observed individual-level feature, generating a mean embedding using the transformed data.

### Claim 3 (depends on 2)

The one or more computer storage media of claim 2, wherein the aggregate data is aggregated over each distinct value of the first observed individual-level feature.

### Claim 4 (depends on 1)

The one or more computer storage media of claim 1, wherein the joint predictive model comprises a Linear Model of Coregionalization model (LMC model), and wherein training the LMC model comprises tuning hyperparameters of the LMC model.

### Claim 5 (depends on 4)

The one or more computer storage media of claim 4, wherein the hyperparameters of the LMC model comprise coefficients of a coregionalization matrix, variables in kernel functions, and a covariance of noise.

### Claim 6 (depends on 1)

The one or more computer storage media of claim 1, wherein the operations further comprise:
generating, from the first individual data instance, transformed data; and
providing the transformed data for the first individual data instance as input to the trained joint predictive model.

### Claim 7 (depends on 1)

The one or more computer storage media of claim 1, wherein the first individual data instance is from the dataset.

### Claim 8

A computer-implemented method comprising:
accessing a dataset stored on a computer storage device, the dataset comprising a plurality of observed individual-level features;
accessing aggregate data associated with the dataset, the aggregate data corresponding to two or more unobserved individual-level features and aggregated over groups defined by a grouping feature of the dataset;
generating, based on the observed individual-level features, a transformed representation of at least a portion of the dataset, the transformed representation encoding relationships among data instances within each group defined by the grouping feature;
training, using the transformed representation and the aggregate data, a joint predictive model configured to learn inter-feature dependencies among the two or more unobserved individual-level features;
applying the trained joint predictive model to input corresponding to a first individual data instance to generate predicted values for each of the two or more unobserved individual-level features; and
augmenting the dataset by associating the predicted values with the first individual data instance.

### Claim 9 (depends on 8)

The computer-implemented method of claim 8, wherein generating the transformed representation of at least the portion of the dataset comprises:
transforming values of the observed individual-level features of each data instance of the dataset to generate transformed data in a higher dimensional space for each data instance; and
for each distinct value of a first observed individual-level feature, generating a mean embedding using the transformed data.

### Claim 10 (depends on 9)

The computer-implemented method of claim 9, wherein the aggregate data is aggregated over each distinct value of the first observed individual-level feature.

### Claim 11 (depends on 8)

The computer-implemented method of claim 8, wherein the joint predictive model comprises a Linear Model of Coregionalization model (LMC model), and wherein training the LMC model comprises tuning hyperparameters of the LMC model.

### Claim 12 (depends on 11)

The computer-implemented method of claim 11, wherein the hyperparameters of the LMC model comprise coefficients of a coregionalization matrix, variables in kernel functions, and a covariance of noise.

### Claim 13 (depends on 8)

The computer-implemented method of claim 8, further comprising:
generating, from the first individual data instance, transformed data; and
providing the transformed data for the first individual data instance as input to the trained joint predictive model.

### Claim 14 (depends on 8)

The computer-implemented method of claim 8, wherein the first individual data instance is from the dataset.

### Claim 15

A system comprising:
one or more processors; and
one or more computer storage media storing computer-useable instructions that, when used by one or more processors, cause the one or more processors to perform operations, the operations comprising:
accessing a dataset stored on a computer storage device, the dataset comprising a plurality of observed individual-level features;
accessing aggregate data associated with the dataset, the aggregate data corresponding to two or more unobserved individual-level features and aggregated over groups defined by a grouping feature of the dataset;
generating, based on the observed individual-level features, a transformed representation of at least a portion of the dataset, the transformed representation encoding relationships among data instances within each group defined by the grouping feature;
training, using the transformed representation and the aggregate data, a joint predictive model configured to learn inter-feature dependencies among the two or more unobserved individual-level features;
applying the trained joint predictive model to input corresponding to a first individual data instance to generate predicted values for each of the two or more unobserved individual-level features; and
augmenting the dataset by associating the predicted values with the first individual data instance.

### Claim 16 (depends on 15)

The system of claim 15, wherein generating the transformed representation of at least the portion of the dataset comprises:
transforming values of the observed individual-level features of each data instance of the dataset to generate transformed data in a higher dimensional space for each data instance; and
for each distinct value of a first observed individual-level feature, generating a mean embedding using the transformed data.

### Claim 17 (depends on 16)

The system of claim 16, wherein the aggregate data is aggregated over each distinct value of the first observed individual-level feature.

### Claim 18 (depends on 15)

The system of claim 15, wherein the joint predictive model comprises a Linear Model of Coregionalization model (LMC model), wherein training the LMC model comprises tuning hyperparameters of the LMC model, and wherein the hyperparameters of the LMC model comprise coefficients of a coregionalization matrix, variables in kernel functions, and a covariance of noise.

### Claim 19 (depends on 15)

The system of claim 15, wherein the operations further comprise:
generating, from the first individual data instance, transformed data; and
providing the transformed data for the first individual data instance as input to the trained joint predictive model.

### Claim 20 (depends on 15)

The system of claim 15, wherein the first individual data instance is from the dataset.
