Topology-aware Reinforcement Feature Space Reconstruction for Graph Data

TL;DR

This work uses topology-aware reinforcement learning to automate and optimize feature space reconstruction for graph data, and combines the extraction of core subgraphs to capture essential structural information with a graph neural network to encode topological features and reduce computing complexity.

Abstract

Feature space is an environment where data points are vectorized to represent the original dataset. Reconstructing a good feature space is essential to augment the AI power of data, improve model generalization, and increase the availability of downstream ML models. Existing literature, such as feature transformation and feature selection, is labor-intensive (e.g., heavy reliance on empirical experience) and mostly designed for tabular data. Moreover, these methods regard data samples as independent, which ignores the unique topological structure when applied to graph data, thus resulting in a suboptimal reconstruction feature space. Can we consider the topological information to automatically reconstruct feature space for graph data without heavy experiential knowledge? To fill this gap, we leverage topology-aware reinforcement learning to automate and optimize feature space reconstruction for graph data. Our approach combines the extraction of core subgraphs to capture essential structural information with a graph neural network (GNN) to encode topological features and reduce computing complexity. Then we introduce three reinforcement agents within a hierarchical structure to systematically generate meaningful features through an iterative process, effectively reconstructing the feature space. This framework provides a principled solution for attributed graph feature space reconstruction. The extensive experiments demonstrate the effectiveness and efficiency of including topological awareness.

Figures (9)

• Figure 1: An Overview of TAR. First, we spot the core subgraph from the original dataset and embed it into a fixed-length embedding to capture the topological information. Second, we group the attributes (a.k.a, features) of the subgraph to enrich the information of features. Finally, we develop a reinforcement system to select features and operations to reconstruct the graph feature space.
• Figure 2: Topological Information Capture with Core Subgraphs and GNNs.
• Figure 3: Feature Grouping over Core Subgraphs. We group the graph attributes (a.k.a., features) using bottom-up hierarchical grouping evaluated by group information distinctness.
• Figure 4: Hierarchical reinforced feature crossing. We develop three reinforcement agents to select feature groups and an operation, and then cross the feature groups to generate new features in an iterative way.
• Figure 5: The impact of subgraph. Subfigure (a), (b), and (c) are F1-score comparisons of three different tasks across all datasets. Subfigure (d), (e), and (f) are training time cost comparisons of three different tasks across all datasets. We report mean values with standard deviation by 10 repeats.