v325: Proceedings of GTML 2025

2026-06-30 · Source: Proceedings of Machine Learning Research · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Mathematics & Computational Sciences · Depth: Expert, quick

Summary

Volume 325 compiles the proceedings from the Geometry, Topology, and Machine Learning Workshop, held from 10-14 November 2025 at the Max Planck Institute for Mathematics in the Sciences in Leipzig, Germany. Edited by Michael Bleher, Freya Jensen, Levin Maier, Diaaeldin Taha, and Anna Wienhard, this volume features a preface, five full papers, and thirteen extended abstracts. Key research areas explored include understanding learning invariance in deep linear networks, developing complete and efficient covariants for 3D point configurations, and applying persistent homology convolutions for histopathology slide classification. Extended abstracts delve into topics such as zigzag persistence in Large Language Model representations and neural responses, the geometry of interbrain networks, and the unification of transformers and convolutional networks as equivariant maps. Other contributions address the geometry of nonlinear reinforcement learning and GNN rewiring.

Key takeaway

For research scientists focused on the mathematical underpinnings of AI, reviewing Volume 325 offers insights into emerging trends. You can explore novel applications of geometry and topology to deep learning, graph neural networks, and large language models. Consider investigating specific papers on learning invariance, persistent homology, or equivariant maps to inform your own theoretical or applied research directions. This collection highlights diverse approaches to complex AI challenges.

Key insights

This volume explores the mathematical foundations and applications of geometry and topology in machine learning.

Topics

• Geometry
• Topology
• Machine Learning
• Deep Learning
• Persistent Homology
• Large Language Models

Best for: AI Scientist, Research Scientist

Related on AIssential

• Machine Learning on Spherical Manifold [R] — Shape correspondence can be robustly computed by minimizing the area of an implicitly encoded mapping in a product space.
• How did Gpt solve the erdos problem? A demonstration — AI's problem-solving success stems from uncovering hidden algebraic structures, not just surface patterns.
• IEEE Transactions on Neural Networks and Learning Systems, Volume 37, Issue 4, April 2026 — The volume showcases diverse neural network and learning system advancements across various applications and theoretical foundations.
• Tracking Representation Dynamics in Large Language Models with Persistent Homology — Persistent homology uncovers internal representation dynamics in LLMs during fine-tuning, showing early topological reorganization and objective-specific trajectories.
• Shape, Symmetries, and Structure: The Changing Role of Mathematics in Machine Learning Research — Mathematics' role in ML is shifting from theoretical guarantees to post-hoc explanations and high-level architectural guidance, leveraging advanced pure math.
• Prerequisites II: Topology - Cristian Bodnar — Topology qualitatively defines proximity and continuity in spaces, abstracting beyond specific distance metrics.

Open in AIssential →

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