Consensus Graph Learning for Multi-View Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-view clustering algorithms face challenges in constructing high-quality similarity graphs due to noise and redundancy in original features, limiting their clustering performance.
Innovation Solution
A consensus graph learning-based multi-view clustering method that learns a consistent similarity graph from a new feature space by obtaining spectral embedding representations, stacking inner products into a third-order tensor, and using low-rank tensor representation learning to filter out noise and redundancy, thereby constructing a robust similarity graph for clustering.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multi-view clustering algorithms directly learn similarity graph from original features, then the process is simple, but the obtained similarity graph is not accurate due to noise and redundant information
Solution Approach 1:
The patent segments the learning process into two distinct stages: first learning spectral embedding representations from each view separately, then learning a consensus similarity graph from these embeddings. This segmentation allows filtering out view-specific noise while preserving meaningful patterns, resolving the contradiction between accuracy and complexity by structuring the learning process hierarchically
Solution Approach 2:
The patent introduces spectral embedding representations as an intermediary between original features and the final similarity graph. These embeddings serve as a mediating representation that filters noise and redundancy while preserving essential clustering information, enabling accurate similarity graph construction without directly processing noisy original features
2Reliability
If spectral embedding representation learning and low-rank tensor representation learning are integrated into unified learning framework, then clustering performance is improved, but computational complexity increases
Solution Approach 1:
The patent merges spectral embedding learning and low-rank tensor representation learning into a unified optimization framework with a single objective function. This combination leverages the strengths of both approaches: spectral embedding captures local structure while low-rank tensor modeling enforces global consistency across views, achieving improved clustering performance through synergistic integration
Solution Approach 2:
The patent transforms the complex unified learning problem into a series of simpler sub-problems by parameterizing the solution and applying alternating optimization. By fixing certain parameters and optimizing others in turns, the method makes the computationally intensive unified framework tractable while maintaining its performance benefits
Data Source
AI summary
A consensus graph learning-based multi-view clustering method includes: S11, inputting an original data matrix to obtain a spectral embedding matrix; S12, calculating a similarity graph matrix and a Laplacian matrix based on the spectral embedding matrix; S13, applying spectral clustering to the calculated similarity graph matrix to obtain spectral embedding representations; S14, stacking inner products of the normalized spectral embedding representations into a third-order tensor and using low-rank tensor representation learning to obtain a consistent distance matrix; S15, integrating spectral embedding representation learning and low-rank tensor representation learning into a unified learning framework to obtain a objective function; S16, solving the obtained objective function through an alternative iterative optimization strategy; S17, constructing a consistent similarity graph based on the solved result; and S18, applying spectral clustering to the consistent similarity graph to obtain a clustering result. A consistent similarity graph for clustering is constructed based on spectral embedding features.