Semi-supervised Learning Model Update via Attribute Distance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning techniques face challenges in generating high-inference accuracy using unlabeled data due to incorrect pseudo label generation and ineffective learning progress.
Innovation Solution
An information processing apparatus and method that acquire labeled and unlabeled data, calculate inter-attribute distances, extract unlabeled data closer to labeled data, and update model parameters using both labeled and extracted unlabeled data to improve learning model accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large amount of labeled training data is used to construct a high-accuracy trained model, then the inference accuracy is improved, but the operating cost increases due to the need to collect and label large numbers of images
Solution Approach 1:
The patent introduces an intermediary selection mechanism based on attribute information distance. Instead of directly using all unlabeled data or requiring extensive manual labeling, the system calculates distances between attribute information of labeled and unlabeled data to selectively choose unlabeled data for training. This intermediary selection process enables effective semi-supervised learning with reduced labeled data while maintaining high inference accuracy.
2Quantity of substance
If unlabeled data is used to reduce operating cost, then the quantity of labeled data required is reduced, but incorrect pseudo label generation occurs and learning progress fails
Solution Approach 1:
The patent applies local quality by selectively using unlabeled data based on their attribute information distance from labeled data. Instead of uniformly processing all unlabeled data, the system identifies and selects unlabeled data with similar attributes (local similarity) to paired labeled data. This selective approach ensures that pseudo labels are generated only for suitable candidates, preventing incorrect label propagation and ensuring reliable learning progress.
3Productivity
If all unlabeled data is used for training, then the productivity of model training is improved, but the manufacturing precision of the learning model decreases due to inclusion of inappropriate data
Solution Approach 1:
The patent changes the selection parameter from binary (labeled/unlabeled) to continuous (distance-based attribute similarity). By introducing attribute information distance as a selection criterion, the system dynamically evaluates and selects unlabeled data based on their similarity to labeled data. This parameter change enables efficient utilization of unlabeled data while maintaining model accuracy, resolving the contradiction between training productivity and manufacturing precision.
Data Source
AI summary
In order to provide a technique for generating a high-inference accuracy learning model in machine learning in which unlabeled data is used, an information processing apparatus includes: an acquiring section configured to acquire labeled data and unlabeled data, the labeled data being image data to which a ground-truth label and attribute information are attached, the unlabeled data being image data to which an attribute information is attached; an inter-attribute distance calculating section configured to calculate a distance between the labeled data and the unlabeled data, the distance being determined by the attribute information; an extracting section configured to extract, from multiple pieces of unlabeled data each being the unlabeled data, unlabeled data smaller in the distance from the labeled data than another unlabeled data; and an updating section configured to update a model parameter with use of the labeled data and the unlabeled data extracted by the extracting section.