Autoencoder Architecture for Semi-Supervised Data Labeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine learning systems face challenges in achieving high classification accuracy due to the limited availability of accurately labeled data, which is time-consuming and costly to produce manually, resulting in a reliance on small amounts of labeled data for training classifiers.
Innovation Solution
A computer-implemented method using an autoencoder architecture that iteratively processes unlabeled data to predict probability distributions, automatically associating labels by updating probabilities based on reconstruction loss, thereby increasing the amount of labeled data available for training classifiers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual methods are used to label previously unlabeled data, then classification accuracy is improved, but time consumption and cost increase significantly
Solution Approach 1:
The system uses semi-supervised learning where the model automatically labels unlabeled data using predictions from labeled data, eliminating the need for continuous manual intervention. The algorithm serves itself by generating labels that can be used to improve its own performance iteratively.
Solution Approach 2:
The patent introduces an intermediary labeling process that uses a combination of labeled data and unlabeled data through semi-supervised learning algorithms. This intermediary step generates predicted labels that bridge the gap between having no labels and requiring full manual labeling, reducing both time and cost while maintaining accuracy.
2Reliability
If manual labeling is performed by experts to ensure high accuracy, then labeled data quality is improved, but development cost increases
Solution Approach 1:
The system applies partial manual labeling by experts to create an initial labeled dataset, which is then used to train the semi-supervised learning model. The model handles the remaining labeling task automatically, performing only the necessary portion of manual work while achieving high data quality through the combination of expert-labeled and model-predicted labels.
3Productivity
If only a small amount of labeled data is used for training, then development time is reduced, but classifier accuracy deteriorates
Solution Approach 1:
The system performs preliminary action by using a small initial labeled dataset to train the semi-supervised learning model. This preliminary training enables the model to generate predictions for unlabeled data, which are then used to expand the labeled dataset before final classifier training, achieving high accuracy without waiting for extensive manual labeling.
Data Source
AI summary
A computer-implemented method, a computing system, and a computer program product, for automatically labeling an amount of unlabeled data for training one or more classifiers of a machine learning system. A method includes iteratively processing unlabeled data items. Receiving an unlabeled data item into each autoencoder in an autoencoder architecture. Each autoencoder processing with a lowest loss of information the unlabeled data item that is likely associated with a label associated with the autoencoder, while processing with a higher loss of information the unlabeled data item that is likely not associated with the label. Predicting, based on loss of information, a probability distribution for the unlabeled data item. Automatically associating the label to the unlabeled data item, based on the label being associated with a highest probability in a peaking probability distribution associated with the unlabeled data item. The autoencoder architecture can include a cloud computing network architecture.