Semantic Metric Learning for Anchor-Based Data Classification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing metric learning methods fail to adequately consider semantic relationships between data clusters, leading to suboptimal classification performance and increased computational complexity, particularly in applications requiring semantic understanding of input data such as scene understanding and emotion analysis.

Innovation Solution

An electronic apparatus and method that utilizes an artificial neural network to train object recognition models by updating anchor points in an embedding space, considering semantic relationships between classes, and optimizing the position of feature points to enhance clustering and reduce computational load.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional metric learning methods are used to classify data, then classification can be performed, but classification accuracy is suboptimal because semantic relationships between data clusters are not adequately considered

Engineering Contradiction:
Improveclassification accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent transforms the classification problem from traditional metric space to semantic embedding space by changing the parameter representation. Semantic embeddings are generated using language models that capture semantic relationships between data points, fundamentally altering how similarity is measured and enabling more accurate classification that understands meaning rather than just numerical proximity

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent introduces semantic embeddings as an intermediary layer between raw data and classification. These embeddings act as a mediator that translates data into a representation where semantic relationships are preserved, allowing the classification algorithm to operate on meaningful representations rather than raw features, thus improving accuracy without proportionally increasing complexity

Inventive Principle:
Principle #24Intermediary (Mediator)

2Productivity

If traditional metric learning methods are used, then classification can be performed, but computational requirements are high due to inadequate optimization

Engineering Contradiction:
Improvecomputational efficiencyVSAvoidclassification accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent performs preliminary action by pre-computing semantic embeddings for all data points before classification. This preprocessing step captures semantic relationships in advance, transforming the classification task into a simpler similarity search problem in embedding space, which significantly reduces computational requirements during actual classification while maintaining high accuracy

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3678064B1Apparatus related to metric-learning-based data classification and method thereof
Publication Date: 2026.03.04 SAMSUNG ELECTRONICS CO LTD
  • EP3678064B1 patent drawingFigure 1
  • EP3678064B1 patent drawingFigure 2
  • EP3678064B1 patent drawingFigure 3

AI summary

The present invention provides artificial intelligence technology which has machine-learning-based information understanding capability, including metric learning providing improved classification performance, classification of an object considering a semantic relationship, understanding of the meaning of a scene based on the metric learning and the classification, and the like. An electronic device according to one embodiment of the present invention comprises a memory in which at least one instruction is stored, and a processor for executing the stored instruction. Here, the processor extracts feature data from training data of a first class, obtains a feature point by mapping the extracted feature data to an embedding space, and makes an artificial neural network learn in a direction for reducing a distance between the obtained feature point and an anchor point.