# US 2024/0202167 A1 · Unsupervised learning from public tabular datasets

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

- Number: US 2024/0202167 A1
- Document ID: US20240202167A1
- Status: Published application
- Kind: A1
- Application number: 18/066,327
- Filed: 2022-12-15
- Published: 2024-06-20
- Earliest claimed: 2022-12-15
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Thanh Lam Hoang; Gabriele Picco; Lam Minh Nguyen; Dzung Tien Phan
- Web page: https://patentgrep.com/patent/US20240202167A1

## CPC

- G06F 16/211: {Schema design and management}
- G06F 16/2282: {Tablespace storage structures; Management thereof}
- G06N 3/045: Combinations of networks
- G06N 3/047: Probabilistic or stochastic networks
- G06N 3/088: Non-supervised learning, e.g. competitive learning
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06N 20/00: Machine learning
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A method, computer program product and system are provided for feature engineering and synthetic data generation. A processor retrieves a plurality of data tables, where the plurality of data tables are heterogeneous in format and content. A processor trains a variational auto-encoder (VAE) model on the plurality of data tables. A processor receives an input data table. A processor generates a synthetic data table based on the input data table and the trained VAE model.

## Claims

### Claim 1

A computer-implemented method for transferable feature engineering and synthetic data generation, the computer-implemented method comprising:
retrieving a plurality of data tables, wherein the plurality of data tables are heterogeneous in format and content;
training a variational auto-encoder (VAE) model on the plurality of data tables;
receiving an input data table; and
generating a synthetic data table based on the input data table and the trained VAE model.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, the computer-implement method further comprising:
generating a list of latent features present in the input data based on the input data table and the trained VAE model.

### Claim 3 (depends on 1)

The computer-implemented method of claim 1, the computer-implement method further comprising:
normalizing the values of the plurality of data tables based on a Gaussian Mixture Model (GMM).

### Claim 4 (depends on 1)

The computer-implemented method of claim 1, the computer-implement method further comprising:
generating signatures for the columns of the plurality of data tables based on one or more of the following: column name, column data type, column categories and metadata regarding the plurality of data tables.

### Claim 5 (depends on 4)

The computer-implemented method of claim 4, wherein training the VAE model is further based on the signatures for the columns of the plurality of data table.

### Claim 6 (depends on 1)

The computer-implemented method of claim 1, wherein training the VAE model further comprises:
supplying at least one table of the plurality of data tables to an input layer of the VAE model;
supplying at least one row of the plurality of data tables to an output layer of the VAE model; and
determining an encoding to generate the at least one row based on the at least one table.

### Claim 7 (depends on 6)

The computer-implemented method of claim 6, wherein the VAE model is trained without supervision.

### Claim 8

A computer program product for transferable feature engineering and synthetic data generation, the computer program product comprising:
one or more computer-readable storage media and program instructions stored on the one or more computer-readable storage media, the program instructions comprising:
program instructions to retrieve a plurality of data tables, wherein the plurality of data tables are heterogeneous in format and content;
program instructions to train a variational auto-encoder (VAE) model on the plurality of data tables;
program instructions to receive an input data table; and
program instructions to generate a synthetic data table based on the input data table and the trained VAE model.

### Claim 9 (depends on 8)

The computer program product of claim 8, the program instructions further comprising:
program instructions to generate a list of latent features present in the input data based on the input data table and the trained VAE model.

### Claim 10 (depends on 8)

The computer program product of claim 8, the program instructions further comprising:
program instructions to normalize the values of the plurality of data tables based on a Gaussian Mixture Model (GMM).

### Claim 11 (depends on 8)

The computer program product of claim 8, the program instructions further comprising:
program instructions to generate signatures for the columns of the plurality of data tables based on one or more of the following: column name, column data type, column categories and metadata regarding the plurality of data tables.

### Claim 12 (depends on 11)

The computer program product of claim 11, wherein program instructions to train the VAE model is further based on the signatures for the columns of the plurality of data table.

### Claim 13 (depends on 8)

The computer program product of claim 8, wherein program instructions to train the VAE model further comprises:
program instructions to supply at least one table of the plurality of data tables to an input layer of the VAE model;
program instructions to supply at least one row of the plurality of data tables to an output layer of the VAE model; and
program instructions to determine an encoding to generate the at least one row based on the at least one table.

### Claim 14 (depends on 13)

The computer program product of claim 13, wherein the VAE model is trained without supervision.

### Claim 15

A computer system for transferable feature engineering and synthetic data generation, the computer system comprising:
one or more computer processors;
one or more computer readable storage media; and
program instructions stored on the computer readable storage media for execution by at least one of the one or more processors, the program instructions comprising:
program instructions to retrieve a plurality of data tables, wherein the plurality of data tables are heterogeneous in format and content;
program instructions to train a variational auto-encoder (VAE) model on the plurality of data tables;
program instructions to receive an input data table; and
program instructions to generate a synthetic data table based on the input data table and the trained VAE model.

### Claim 16 (depends on 15)

The computer system of claim 15, the program instructions further comprising:
program instructions to generate a list of latent features present in the input data based on the input data table and the trained VAE model.

### Claim 17 (depends on 15)

The computer system of claim 15, the program instructions further comprising:
program instructions to normalize the values of the plurality of data tables based on a Gaussian Mixture Model (GMM).

### Claim 18 (depends on 15)

The computer system of claim 15, the program instructions further comprising:
program instructions to generate signatures for the columns of the plurality of data tables based on one or more of the following: column name, column data type, column categories and metadata regarding the plurality of data tables.

### Claim 19 (depends on 18)

The computer system of claim 18, wherein program instructions to train the VAE model is further based on the signatures for the columns of the plurality of data table.

### Claim 20 (depends on 15)

The computer system of claim 15, wherein program instructions to train the VAE model further comprises:
program instructions to supply at least one table of the plurality of data tables to an input layer of the VAE model;
program instructions to supply at least one row of the plurality of data tables to an output layer of the VAE model; and
program instructions to determine an encoding to generate the at least one row based on the at least one table.
