Data Augmentation Program for Natural Training Data

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing data augmentation methods often generate unnatural data by violating correlations between attributes or data distribution, leading to inaccurate training data for machine learning models.

Innovation Solution

A data augmentation program that generates artificial data based on existing data and external statistical information, using a base artificial data generation unit to create data belonging to minority classes while ensuring attribute values align with real-world correlations, and an optimization unit to minimize errors and unnatural data generation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If existing data augmentation methods are used to generate artificial data, then the quantity of training data is increased, but the quality of training data deteriorates due to unnatural data generation that violates attribute correlations

Engineering Contradiction:

Improvequantity of training dataVSAvoidquality of training data

Core Design Contradiction:

Quantity of substanceVSManufacturing precision

Solution Approach 1:

The patent changes the parameters used in data generation by incorporating statistical information about attribute correlations and relationships. Instead of simple random generation, the system adjusts generation parameters based on learned statistical patterns from existing data, ensuring that generated data maintains realistic attribute relationships while increasing data quantity.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent implements a feedback mechanism where the system evaluates generated artificial data against statistical information and attribute correlation rules. If generated data violates established correlations or relationships, the system adjusts the generation process accordingly, creating a closed-loop system that continuously improves data quality while maintaining quantity.

Inventive Principle:

Principle #23Feedback

2Productivity

If artificial data is generated without considering attribute correlations, then data generation speed is improved, but data naturalness deteriorates

Engineering Contradiction:

Improvedata generation speedVSAvoiddata naturalness

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent performs preliminary analysis to extract statistical information and attribute correlations from existing data before the actual data generation process. By pre-computing these relationships and storing them as reference information, the system can quickly generate natural-looking data during the augmentation phase without performing complex real-time calculations, thus maintaining both speed and naturalness.

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If statistical information and attribute relationships are incorporated into data generation, then data accuracy is improved, but processing complexity increases

Engineering Contradiction:

Improvedata accuracyVSAvoidprocessing complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the data generation process into distinct modules: one for extracting statistical information from existing data, another for learning attribute correlations, and a final module for generating artificial data using these learned patterns. This segmentation allows each module to specialize in a specific task, improving overall accuracy while making the complex process more manageable and efficient through modular design.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS20240070229A1Computer-readable recording medium storing data augmentation program, data augmentation method, and data augmentation apparatus

Publication Date: 2024.02.29 FUJITSU LTD

AI summary

A non-transitory computer-readable recording medium stores a data augmentation program for causing a computer to execute processing including: acquiring a first plurality of pieces of data and statistical information regarding a plurality of attributes included in each of the first plurality of pieces of data; specifying a relationship between the attributes in the first plurality of pieces of data based on values of the plurality of attributes included in each of the first plurality of pieces of data; and generating data based on the first plurality of pieces of data, the statistical information, and the relationship between the attributes in the first plurality of pieces of data.