Urban Graph Node Feature Updating via Spatial Partitioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Urban graphs exhibit spatial heterophily, which limits the performance of general graph neural networks in accurately representing urban entities due to dissimilar nodes being connected, leading to poor quality of target feature data during node feature updating.
Innovation Solution
The method involves partitioning a target region into sub-regions based on spatial positions, generating regional features, and updating node features using relation information to alleviate spatial heterophily, ensuring that spatial information is fully considered for generating target feature data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If general graph neural networks are used to update node features in urban graphs, then the model can process graph data, but the quality of target feature data deteriorates due to spatial heterophily causing dissimilar nodes to be connected
Solution Approach 1:
The patent partitions the target region into multiple sub-regions based on spatial positions, and further divides sub-regions into groups. This segmentation allows the model to process spatially heterogeneous nodes separately, preventing dissimilar nodes from being incorrectly aggregated and improving the quality of target feature data.
Solution Approach 2:
The patent applies different processing strategies to different spatial regions by generating relation information specific to each sub-region and group. This local quality approach ensures that spatial relationships are preserved and utilized appropriately for each region, addressing the spatial heterophily problem while maintaining adaptability.
2Loss of information
If node features are updated using all neighboring nodes, then the feature aggregation is comprehensive, but spatial information is lost due to treating all neighbors uniformly
Solution Approach 1:
By partitioning the target region into sub-regions and groups, the patent preserves spatial information by maintaining the spatial relationships among nodes. Each sub-region and group processes nodes with similar spatial characteristics, preventing loss of spatial information while managing complexity through structured organization.
Solution Approach 2:
The patent generates relation information specific to each sub-region and group, applying local quality to preserve spatial characteristics. This approach maintains spatial information by treating different spatial regions differently, while the systematic structure manages the complexity of the feature updating process.
3Reliability
If the target region is partitioned into sub-regions and relation information is generated, then spatial heterophily is alleviated, but the computational complexity increases
Solution Approach 1:
The patent partitions the target region into sub-regions and groups to alleviate spatial heterophily, improving the representation of urban entities. The segmentation approach manages computational complexity by organizing nodes into structured groups that can be processed systematically, rather than handling all nodes individually.
Solution Approach 2:
The patent merges nodes within the same group to generate representative features, reducing computational complexity. By combining nodes with similar spatial characteristics into groups and processing them together, the model improves representation quality while managing the complexity of data processing through efficient aggregation.
Data Source
AI summary
A data updating method, a model training method and related devices are provided. The method includes obtaining urban graph data in a preset region, the urban graph data including a node set including central nodes, an edge set and a feature set, the edge set including neighborhoods corresponding to the central nodes, the neighborhoods including other nodes possessing connecting edges with the central nodes, the neighborhoods corresponding to a target region, and the feature set including node features of the nodes in the node set; partitioning the target region into at least two sub-regions to obtain a region partition set; aggregating the node features corresponding to all nodes located within the same sub-region to obtain the regional features of each of the sub-regions; updating the node features of the central node based on the regional features of the sub-regions in the region partition set to obtain target feature data.