Neural Network Input Node Mapping via Segmented Conversion Information

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing machine learning techniques using neural networks face challenges in accurately determining the correspondence between input values and nodes, leading to suboptimal learning accuracy, especially when relationships with specific items or objects are not considered.

Innovation Solution

A method that generates common and individual conversion information to refine the correspondence between input values and nodes, based on similarities between test and collation data, allowing for updated collation data and neural network training that considers specific item relationships.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional machine learning techniques are used to determine correspondence between input values and nodes, then the processing is simple, but the learning accuracy is insufficient

Engineering Contradiction:

Improvelearning accuracyVSAvoidcorrespondence determination complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The correspondence determination is segmented into two independent parts: common conversion information that is universally applied across all data, and individual conversion information that is specific to each data item. This segmentation allows the system to capture both general patterns and specific relationships, thereby improving learning accuracy without overwhelming complexity in a single conversion mechanism.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces a new dimension to the correspondence determination by incorporating conversion information as an additional layer between input values and nodes. Instead of direct mapping, the system uses conversion information (both common and individual) as an intermediary dimension that enriches the correspondence relationship, enabling more accurate learning while maintaining structured processing.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Measurement precision

If individual conversion information is generated for each data item, then the correspondence accuracy improves, but the processing time increases

Engineering Contradiction:

Improvecorrespondence accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

By segmenting the conversion information into common and individual components, the system can pre-compute and cache the common conversion information that applies to all data items, while only computing individual conversion information when needed for specific data items. This reduces overall processing time while maintaining high correspondence accuracy through the individual conversion component.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The common conversion information is generated in advance before processing individual data items. This preliminary action allows the system to reuse the common conversion information for all data items without recomputing it, significantly reducing processing time while still achieving accurate correspondence through the combination of common and individual conversion information.

Inventive Principle:

Principle #10Preliminary action

3Reliability

If collation data is updated frequently to improve learning, then the learning performance improves, but the system stability decreases

Engineering Contradiction:

Improvelearning performanceVSAvoidcollation data stability

Core Design Contradiction:

ReliabilityVSStability of the object's composition

Solution Approach 1:

The collation data is segmented into common conversion information and individual conversion information. The common conversion information serves as a stable foundation that is updated less frequently and applies universally, while individual conversion information can be updated more frequently for specific data items. This segmentation allows the system to maintain overall stability through the common component while adapting to new learning requirements through the individual component.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Different parts of the collation data have different update frequencies and stability characteristics. The common conversion information maintains high stability and is updated rarely, while the individual conversion information can be updated more frequently to capture specific relationships. This local quality differentiation allows the system to balance learning performance improvement with overall system stability.

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS20210295156A1Non-transitory computer-readable storage medium for storing machine-learning program, machine-learning method, and information processing device

Publication Date: 2021.09.23 FUJITSU LTD

AI summary

A method includes: generating common information to be commonly applied to plural input data each including a combination of a value of each item and an input value in association with one or more items, the common information being for converting a correspondence between each input value and each input node in a machine learner in a case of inputting the plural input data to the machine learner; generating individual information to be individually applied to each input data, the individual information being for converting the correspondence, in association with a remaining item excluding the one or more items, by using a similarity between test data and collation data obtained by converting the correspondence; generating converted data obtained by converting the correspondence by using the generated common conversion information and the generated individual conversion information; and updating the collation data and the machine learner by using the generated converted data.