Universally Invariant Learning in Equivariant GNNs

2025-10-15

AI Summary

This work addresses the challenge of designing equivariant graph neural networks (GNNs) that simultaneously achieve universal approximation, computational efficiency, and polynomial-time solvability. We propose an efficient and complete construction framework grounded in geometric graphs: by introducing canonical forms and full-rank steerable basis sets, we reduce high-order tensor operations to scalar function modeling—enabling universal approximation of arbitrary equivariant functions via linear combinations of a canonical scalar network and the basis set. Unlike conventional higher-order tensor-based methods (e.g., EGNN, TFN), our approach avoids exponential complexity, achieving polynomial-time algorithmic complexity. Experiments demonstrate state-of-the-art performance on molecular dynamics and equivariant regression tasks using only 2–3 layers, with up to 5.3× speedup and strong expressivity.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines

Abstract

Equivariant Graph Neural Networks (GNNs) have demonstrated significant success across various applications. To achieve completeness -- that is, the universal approximation property over the space of equivariant functions -- the network must effectively capture the intricate multi-body interactions among different nodes. Prior methods attain this via deeper architectures, augmented body orders, or increased degrees of steerable features, often at high computational cost and without polynomial-time solutions. In this work, we present a theoretically grounded framework for constructing complete equivariant GNNs that is both efficient and practical. We prove that a complete equivariant GNN can be achieved through two key components: 1) a complete scalar function, referred to as the canonical form of the geometric graph; and 2) a full-rank steerable basis set. Leveraging this finding, we propose an efficient algorithm for constructing complete equivariant GNNs based on two common models: EGNN and TFN. Empirical results demonstrate that our model demonstrates superior completeness and excellent performance with only a few layers, thereby significantly reducing computational overhead while maintaining strong practical efficacy.

Problem

• Achieving universal approximation in equivariant GNNs efficiently
• Reducing computational costs while maintaining completeness in GNNs
• Constructing complete equivariant networks with minimal layers practically

Innovation

• Complete scalar function as canonical geometric graph form
• Full-rank steerable basis set for equivariant representation
• Efficient algorithm enabling shallow complete equivariant GNNs