Smell Data Label Candidate Generation for Flexible ML Training

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing technologies for machine learning are limited by the use of pre-defined sensory evaluation choices as correct answer labels, preventing the use of desired correct answer labels for training data generation.

Innovation Solution

A training data generation device and method that acquires smell data, generates label candidates based on associated information, receives user selection of labels, and generates training data using the selected labels and smell data, allowing for flexible and desired correct answer labels in machine learning applications.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If pre-defined sensory evaluation choices are used as correct answer labels, then the machine learning system has a fixed and simple label structure, but it cannot perform machine learning using desired correct answer labels

Engineering Contradiction:

Improveflexibility of correct answer labelsVSAvoidcomplexity of label generation system

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system performs preliminary actions by pre-generating multiple label candidates based on smell data characteristics before the actual labeling process. This allows the system to have flexible correct answer labels while maintaining a structured approach through pre-computed candidates.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

Label candidates serve as an intermediary between the smell data and the final correct answer label. The system generates multiple candidate labels that mediate between the raw smell data and the desired correct answer, enabling flexibility while maintaining system structure.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If multiple label candidates are generated and user selection is required, then desired correct answer labels can be used for training, but the operation process becomes more complex

Engineering Contradiction:

Improveaccuracy of training data labelsVSAvoidsimplicity of label selection process

Core Design Contradiction:

ReliabilityVSEase of operation

Solution Approach 1:

The system applies local quality by presenting multiple label candidates with different qualities or confidence levels. Users can select from locally optimized candidates that are most relevant to their needs, improving label accuracy while keeping the interface manageable through focused candidate presentation.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The system generates excessive label candidates beyond what might be strictly necessary, allowing users to select from a comprehensive set of options. This partial action approach ensures that the desired correct answer label is available among the candidates while maintaining user control over the final selection.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20230061026A1Training data generation device, training data generation method, and programrecording medium

Publication Date: 2023.03.02 NEC CORP

AI summary

A training data generation device includes a label candidate generation unit, a reception unit, and a training data generation uni. The acquisition unit is configured to acquire smell data and information pertaining to the smell data. The label candidate generation unit which generates label candidates on the basis of the information pertaining to the smell data; an output unit which outputs the generated label candidates. The reception unit is configured to receive selection of a label from the output label candidates. The training data generation unit which generates training data from the selected label and the smell data.