Neural Network Graph Processing with Assumed Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional neural network technologies that reflect graph structures struggle to handle large-scale, diverse, and variable data, particularly in applications like image recognition and social infrastructure analysis, where they fail to efficiently process complex graph structures and adapt to changes.
Innovation Solution
An information processing device and method that generates a neural network based on graph structure data by introducing assumed nodes and edges, using a propagation matrix and coefficients to propagate feature amounts across layers, allowing for flexible adaptation and accurate prediction of states in complex networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional neural network technology is used to process graph structure data, then basic graph processing is possible, but it cannot efficiently handle large-scale, diverse, and variable data
Solution Approach 1:
The patent segments the graph structure data into nodes and edges as separate entities. Nodes represent entities (e.g., power stations, substations) while edges represent relationships (e.g., transmission lines). This segmentation allows the neural network to process each component independently and combine results, enabling efficient handling of large-scale graphs while maintaining adaptability to different data types and scales.
2Adaptability or versatility
If conventional neural network architecture is applied to graph structures, then some graph processing capability is achieved, but processing load increases significantly
Solution Approach 1:
The patent implements a dynamic neural network architecture where the network structure adapts to the input graph data. The network dynamically determines which nodes and edges to process based on the specific problem and data characteristics, rather than processing the entire graph uniformly. This dynamic approach reduces computational load by focusing resources on relevant portions of the graph while maintaining the capability to handle diverse graph structures.
3Productivity
If conventional neural networks process graph data with fixed structures, then processing is efficient, but adaptability to changes in target data is poor
Solution Approach 1:
The patent employs parameter changes to enable adaptability. The neural network uses learnable parameters (weights and biases) that are optimized during training to handle specific graph structures. When presented with new or changed graph data, the network can adjust these parameters through continued learning or by selecting from pre-trained parameter sets, thereby adapting to different data characteristics while maintaining efficient processing through the established network architecture.
Data Source
AI summary
An information processing device of embodiments includes a data acquirer and a network processor. The data acquirer is configured to acquire graph structure data that includes a plurality of real nodes and one or more real edges connecting two of the plurality of real nodes. The network processor is configured to execute processing of propagating a feature amount of a k−1th layer of each of a plurality of assumed nodes that include the plurality of real nodes and the one or more real edges at least to a feature amount of a kth layer of another assumed node in a connection relationship with each of the assumed nodes in a neural network on the basis of the graph structure data acquired by the data acquirer. k is a natural number equal to or more than 1.