Manifold Learning Mapping Model for Industrial Data Dimensionality Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional statistical analytics and machine learning algorithms face inefficiencies and high computational costs when dealing with high-dimensional data, particularly in image sensor data, due to the curse of dimensionality, which hinders effective analysis and processing in industrial applications.
Innovation Solution
The proposed method involves using a computer-implemented system that applies a mapping model to reduce dimensionality through manifold learning, determining if the model needs updating, and if so, learning new parameters; otherwise, applying a classification model for prediction, enabling efficient processing and classification of sensed data points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional statistical analytics and machine learning algorithms are applied to high-dimensional data, then analysis capability is maintained, but computational efficiency deteriorates and processing time increases
Solution Approach 1:
The patent segments the high-dimensional data processing task into two distinct stages: (1) dimensionality reduction stage using manifold learning to project data onto a lower-dimensional manifold, and (2) classification stage using a trained classifier on the reduced-dimensional data. This segmentation resolves the contradiction by handling the computationally intensive dimensionality reduction separately from the classification task, improving overall efficiency.
Solution Approach 2:
The patent applies manifold learning to transform data from high-dimensional space into a lower-dimensional embedding space while preserving the intrinsic manifold structure. This dimensionality change reduces the computational burden of subsequent classification operations, directly addressing the efficiency-time tradeoff by operating in a reduced-dimensional space.
2Productivity
If manifold learning is applied to reduce dimensionality, then computational efficiency improves, but model accuracy may deteriorate due to information loss
Solution Approach 1:
The patent performs preliminary action by training the classifier on the reduced-dimensional embeddings before deploying the system. This pre-training on manifold-reduced data ensures that the classifier learns the decision boundaries in the lower-dimensional space, maintaining accuracy while benefiting from the efficiency gains of dimensionality reduction.
Solution Approach 2:
The patent incorporates a feedback mechanism that monitors classification performance and triggers re-training of the manifold learning model when accuracy degradation is detected. This feedback loop ensures that the dimensionality reduction process maintains classification accuracy by adapting to changes in the data distribution or manifold structure.
Data Source
AI summary
Method and system that includes receiving an sensed data point from an industrial process; applying a mapping model to map the sensed data point to a respective embedding that has reduced dimensionality relative to the sensed data point; determining, based on a comparison of the respective embedding to prior embeddings, if the mapping model needs to be updated or not. When the mapping model needs to be updated, applying manifold learning to learn an updated set of model parameters for the mapping model. When the mapping model does not need to be updated, applying a classification model to the respective embedding to predict a classification for the sensed data point.