v228: Proceedings of NeuReps 2023

2026-06-10 · Source: Proceedings of Machine Learning Research · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Expert, short

Summary

Volume 228 presents the proceedings of the 2nd NeurIPS Workshop on Symmetry and Geometry in Neural Representations, held on December 16, 2023, in New Orleans. This collection features research applying advanced mathematical and geometric principles to enhance neural network architectures and understanding. Key contributions include novel approaches like Sheaf-based Positional Encodings for Graph Neural Networks, Fast Temporal Wavelet Graph Neural Networks, and Homological Convolutional Neural Networks. The papers also explore topics such as invariant networks, internal representations of vision models, dynamics of neuronal systems, and the geometry of learned knowledge in deep reinforcement learning agents. This volume underscores the critical role of symmetry and geometry in advancing representation learning, model interpretability, and the development of robust AI systems.

Key takeaway

The NeurIPS Workshop on Symmetry and Geometry in Neural Representations (Volume 228) presents advancements in leveraging geometric and symmetric principles to enhance neural network design and understanding. Papers introduce novel techniques like Sheaf-based Positional Encodings for GNNs and Homological CNNs, alongside analyses of internal representations in vision models and deep RL agents. This collection offers critical insights for AI researchers and practitioners aiming to improve model expressivity, robustness, and interpretability across diverse machine learning applications.

Topics

• Neural Representations
• Graph Neural Networks
• Symmetry and Invariance
• Geometric Deep Learning
• Topological Data Analysis

Best for: AI Scientist, Research Scientist

Related on AIssential

• v197: Proceedings of NeurReps Workshop 2022 — Volume 197 compiles the proceedings of the 1st NeurIPS Workshop on Symmetry and Geometry in Neural Representations, held on December 3, 2022, in New Orleans, USA.
• v196 — Volume 196 of the Proceedings of Machine Learning Research (PMLR) compiles papers from the "Topological, Algebraic and Geometric Learning Workshops 2022", held virtually from July 25-22, 2022.
• Neural Networks on Symmetric Spaces of Noncompact Type — A unified point-to-hyperplane distance enables novel neural network architectures on noncompact symmetric spaces.
• Representation Gap: Explaining the Unreasonable Effectiveness of Neural Networks from a Geometric Perspective — Neural network generalization is geometrically characterized by the Representation Gap, scaling with intrinsic data dimension and model equivariance.
• Social-JEPA: Emergent Geometric Isomorphism — Independently trained JEPA world models spontaneously develop linearly alignable latent spaces, enabling efficient knowledge transfer.
• Which Way Does It Go? — Symmetric geometric models are insufficient for systems where direction, order, or causality are fundamental.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Proceedings of Machine Learning Research.