CellCLAT: Preserving Topology and Trimming Redundancy in Self-Supervised Cellular Contrastive Learning

2025-05-27

AI Summary

To address two key challenges in self-supervised topological representation learning on cell complexes—degraded higher-order structural fidelity and semantic redundancy interference—this paper proposes a novel contrastive learning framework. Methodologically: (i) it introduces parameter-perturbed topological-preserving augmentation, explicitly respecting the combinatorial constraints of cell complexes; (ii) it incorporates a two-level meta-learning-driven cell pruning mechanism that adaptively removes redundant topological units via gradient masking. Theoretically, the framework guarantees robustness of augmentations and convergence of optimization. Empirically, it significantly outperforms existing self-supervised graph learning baselines across multiple topology-aware tasks. Notably, it achieves the first robust unsupervised representation learning for cell-complex-level higher-order interactions. This work establishes a new paradigm for higher-order structural modeling—one that is provably sound, controllably optimized, and scalable.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Computer Vision: Representation Learning for Vision
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Self-supervised topological deep learning (TDL) represents a nascent but underexplored area with significant potential for modeling higher-order interactions in simplicial complexes and cellular complexes to derive representations of unlabeled graphs. Compared to simplicial complexes, cellular complexes exhibit greater expressive power. However, the advancement in self-supervised learning for cellular TDL is largely hindered by two core challenges: extit{extrinsic structural constraints} inherent to cellular complexes, and intrinsic semantic redundancy in cellular representations. The first challenge highlights that traditional graph augmentation techniques may compromise the integrity of higher-order cellular interactions, while the second underscores that topological redundancy in cellular complexes potentially diminish task-relevant information. To address these issues, we introduce Cellular Complex Contrastive Learning with Adaptive Trimming (CellCLAT), a twofold framework designed to adhere to the combinatorial constraints of cellular complexes while mitigating informational redundancy. Specifically, we propose a parameter perturbation-based augmentation method that injects controlled noise into cellular interactions without altering the underlying cellular structures, thereby preserving cellular topology during contrastive learning. Additionally, a cellular trimming scheduler is employed to mask gradient contributions from task-irrelevant cells through a bi-level meta-learning approach, effectively removing redundant topological elements while maintaining critical higher-order semantics. We provide theoretical justification and empirical validation to demonstrate that CellCLAT achieves substantial improvements over existing self-supervised graph learning methods, marking a significant attempt in this domain.

Problem

• Preserving cellular topology in self-supervised contrastive learning
• Reducing semantic redundancy in cellular complex representations
• Adapting graph augmentation for higher-order cellular interactions

Innovation

• Parameter perturbation preserves cellular topology
• Cellular trimming scheduler removes redundant elements
• Bi-level meta-learning maintains critical semantics