Knowledge Graph Completion Using Optical Interference Scoring

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing knowledge graph completion methods struggle with adapting symmetrical relations and overlook semantic differentials between entities, leading to inaccurate predictions and incomplete information.

Innovation Solution

An interference-based method using optical interference and superposition principles to represent relations as modulators, combined with Bernoulli probability distribution for negative sampling, to construct a knowledge graph completion model that differentiates between positives and negatives.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If rotation models like RotatE are used to fit triples in a rotating way, then the defect of adapting symmetrical relations is overcome, but phase matching is mainly based and semantic differential of entities about different relations is overlooked, leading to defective prediction

Engineering Contradiction:

Improveadaptation to symmetrical relationsVSAvoidprediction accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent introduces amplitude modulation to break the symmetry of phase-only matching in RotatE. By applying different amplitude modulations to head and tail entities based on their respective relations, the model can distinguish semantic differentials while maintaining symmetrical relation handling capability. This resolves the contradiction by adding asymmetrical amplitude modulation to the previously symmetrical phase matching mechanism.

Inventive Principle:

Principle #4Asymmetry

Solution Approach 2:

The patent applies local quality by using relation-specific amplitude modulation parameters for different entities. Each entity receives customized amplitude modulation based on its local relation context, allowing the model to capture semantic differentials specific to each entity-relation pair while maintaining the global rotational structure for symmetrical relations.

Inventive Principle:

Principle #3Local quality

2Adaptability or versatility

If self-adversarial sampling is used in RotatE, then a better example of sampling is provided, but false negatives are overweighted, making prediction less accurate

Engineering Contradiction:

Improvesampling qualityVSAvoidprediction accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent changes the sampling parameters by introducing Bernoulli probability distribution with dynamically adjusted probabilities. Instead of uniform or fixed adversarial sampling, the sampling probability varies based on the luminous intensity scores, allowing the model to adaptively balance positive and negative sample weights during training, thus preventing false negatives from being overweighted.

Inventive Principle:

Principle #35Parameter changes

3Ease of manufacture

If distance models like TransE are used for representation learning, then the mainstream approach is followed, but symmetrical relations cannot be adapted properly

Engineering Contradiction:

Improvemodel implementation simplicityVSAvoidsymmetrical relation adaptation

Core Design Contradiction:

Ease of manufactureVSAdaptability or versatility

Solution Approach 1:

The patent substitutes the translational mechanism of TransE with a rotational mechanism inspired by optical interference. By representing entities as complex numbers and relations as rotations in the complex plane, the model naturally handles symmetrical relations through rotational symmetry while maintaining mathematical tractability similar to the original TransE approach.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

Improves prediction accuracy by effectively completing missing triples in knowledge graphs, enhancing model generalization and accuracy through enhanced negative sampling and luminous intensity scoring.

Implementation Method 1

based on optical interference and superposition principles, constructing a score function from data of superposed luminous intensities

Methodology Applied

Scientific EffectOptical interference: Interference

Implementation Method 2

mirroring the triples in the knowledge graph to a process of superposition of the luminous intensities

Methodology Applied

Scientific EffectSuperposition principle:

Data Source

PatentUS12541692B2Interference-based method for knowledge graph completion and system thereof

Publication Date: 2026.02.03 HUAZHONG UNIV OF SCI & TECH

AI summary

The present invention relates to an interference-based method for knowledge graph completion and system thereof, wherein the method at least comprises: when performing sampling on a knowledge graph, constructing a knowledge graph completion model; performing model training and performance evaluation on the knowledge graph completion model; and performing prediction on missing elements of incomplete triples in the knowledge graph; the knowledge graph completion model is constructed through: based on optical interference and superposition principles, constructing a score function from data of superposed luminous intensities, mirroring the triples in the knowledge graph to a process of superposition of the luminous intensities, and differentiating between positives and negatives obtained during the sampling of the knowledge graph.