Link Prediction in Multiplex Networks via Layer Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for link prediction in multiplex networks face challenges when dealing with large numbers of layers, as they may incorporate irrelevant information and fail to utilize label information effectively, leading to suboptimal performance.
Innovation Solution
A graph neural network framework is employed to identify structural node labels and determine common and individual embeddings across layers, using a k-nearest approach to select a subset of layers for link prediction, and performing feed-forward networks within the selected subset, while utilizing self-attention mechanisms to combine information from different layers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all layers in a large multiplex network are used for link prediction, then more information is available for prediction, but computational complexity increases and irrelevant information is incorporated
Solution Approach 1:
The patent segments the large multiplex network into multiple layers and further segments the processing by identifying and selecting only a subset of relevant layers for link prediction. This segmentation allows the system to handle large networks by processing manageable portions rather than all layers simultaneously, reducing computational complexity while maintaining prediction accuracy.
Solution Approach 2:
The patent extracts and identifies a subset of relevant layers from the complete multiplex network using graph neural networks to determine which layers contain meaningful information for link prediction. By taking out only the necessary layers and discarding irrelevant ones, the system reduces computational burden while preserving prediction reliability.
2Reliability
If all layers are processed for link prediction, then comprehensive information is utilized, but the model incorporates irrelevant information that reduces performance
Solution Approach 1:
The patent applies local quality by treating different layers differently based on their relevance to the specific link prediction task. Instead of uniformly processing all layers, the system identifies layers with local quality (relevance) and focuses computational resources on those specific layers, while ignoring layers that do not contribute meaningful information for the prediction task.
Solution Approach 2:
The patent implements partial action by processing only a subset of layers rather than all layers in the multiplex network. This selective processing avoids the excessive action of incorporating all available information, including irrelevant data, while still capturing the essential information needed for accurate link prediction.
3Ease of manufacture
If conventional embedding approaches are used in large multiplex networks, then implementation is straightforward, but performance degrades due to irrelevant information incorporation
Solution Approach 1:
The patent applies preliminary action by using graph neural networks to pre-identify and select relevant layers before performing the actual link prediction. This preliminary step of layer selection ensures that only meaningful information is incorporated into the embedding process, preventing performance degradation from irrelevant data while maintaining implementation feasibility.
Solution Approach 2:
The patent introduces graph neural networks as an intermediary between the raw multiplex network data and the link prediction model. This intermediary layer performs the function of filtering and selecting relevant information, mediating between the complexity of the full network and the requirements of the prediction task, thereby improving performance without excessive implementation complexity.
Data Source
AI summary
A method and a system for using a graph neural network framework to implement a link prediction in a multiplex network environment is provided. The method includes: identifying a plurality of layers of a multiplex network, each respective layer including a respective plurality of nodes; for each node included in at least a first layer, providing, by a structural node label and determining a common embedding across all of the plurality of layers and an individual embedding for each individual layer; using a k-nearest approach to select a subset of the plurality of layers for performing link prediction with respect to each layer based on the determined embeddings; and performing a link prediction by determining a respective feed-forward network with respect to each layer included in the selected subset.