Learnijoy
ResearchAI ResearchAI Engineering & DevTools

Phase-Equivariant Learning Captures Cardiac Cyclicity for Diagnostics.

Blaise Delaney, Dominic Dootson, Juan Jose Juan Castella, Salil Patel, Andrew Pfaff, Yuji Xing, Jonny Hancox, Karin Sevegnani· August 24, 2026 View original

Key takeaways

  • Winder uses phase-equivariant self-supervised learning to capture cardiac cyclicity.
  • It organizes representations into phase-invariant and phase-rotating subspaces.
  • The method achieves high diagnostic accuracy with a significantly smaller parameter footprint.
  • Its latent geometry is legible and directly tied to physiological quantities.

Who benefits

HealthcareMedical DevicesWearable TechnologySports ScienceAI Development

Summary

Researchers introduce Winder, a self-supervised learning architecture that explicitly encodes cardiac-phase symmetry to organize representations into phase-invariant coordinates and phase-rotating harmonic subspaces. This method achieves diagnostic accuracy comparable to state-of-the-art self-supervised methods with a significantly smaller parameter footprint, demonstrating legible and parameter-efficient latent geometry tied to physiological quantities.

The cyclic nature of physiological processes, particularly the cardiac cycle, offers a valuable prior for self-supervised representation learning. This research leverages this inherent symmetry by deriving a phase-equivariant self-supervised objective and introducing a joint-embedding architecture named Winder. Winder is designed to organize learned representations into both phase-invariant coordinates and phase-rotating harmonic subspaces. A key innovation of Winder is its fixed and closed-form transport operator, which is derived directly from the cycle's geometry rather than being learned. This design choice means the operator adds no additional parameters to the model, contributing to its efficiency. Evaluated on the PTB-XL dataset using a frozen linear-probe protocol, Winder achieves diagnostic accuracy within the range of state-of-the-art self-supervised methods, but with a significantly smaller parameter footprint of approximately 1 million parameters. The method also yields a latent geometry that is not only legible but also directly linked to a measurable physiological quantity, demonstrating the benefits of explicitly encoding cardiac-phase symmetry for preserving diagnostically useful information.

Why it matters

This advancement provides a more efficient and interpretable way to learn representations from physiological signals, which can lead to more accurate, resource-light, and explainable AI diagnostic tools in healthcare.

How to implement this in your domain

  1. 1Explore Winder's architecture for developing self-supervised learning models on other cyclic physiological data (e.g., respiratory, neurological).
  2. 2Integrate phase-equivariant principles into existing medical AI diagnostic pipelines for improved efficiency and interpretability.
  3. 3Benchmark Winder's performance and parameter efficiency against current state-of-the-art methods in cardiac diagnostics.
  4. 4Collaborate with medical professionals to validate the clinical utility and interpretability of the latent representations.
  5. 5Investigate adapting the fixed transport operator concept to other domains with known cyclic or symmetric structures.

Original post by Blaise Delaney, Dominic Dootson, Juan Jose Juan Castella, Salil Patel, Andrew Pfaff, Yuji Xing, Jonny Hancox, Karin Sevegnani

"arXiv:2608.21147v1 Announce Type: new Abstract: The cyclic structure of physiological processes offers a natural prior for self-supervised representation learning, and the cardiac cycle provides a particularly well-defined setting in which to exploit it. We derive a phase-equivar…"

View on X

Originally posted by Blaise Delaney, Dominic Dootson, Juan Jose Juan Castella, Salil Patel, Andrew Pfaff, Yuji Xing, Jonny Hancox, Karin Sevegnani on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses

More in AI Research

AI ResearchAI Engineering & DevTools

Harmony Improves Protein-Ligand Flexible Docking with Torsional Diffusion

Researchers introduce Harmony, a harmonic torsional diffusion framework for flexible protein-ligand docking that explicitly accounts for the periodic geometry of angular variables. This method improves ligand pose accuracy and pocket all-atom reconstruction on benchmarks like PDBBind and enhances the physical validity of generated complexes on PoseBusters.

Maksim Zhdanov, Pavel Strashnov, Vladislav KurenkovAug 24, 2026
AI Engineering & DevToolsAI Research

Multilingual Verifier Bias Impacts RLVR in LLM Mathematical Reasoning

A study reveals that exact-match verifiers in Reinforcement Learning with Verifiable Rewards (RLVR) for Large Language Models (LLMs) exhibit significant language-dependent false-negative reward noise in multilingual mathematical reasoning. This bias, particularly pronounced in Japanese, stems from format and script variations, highlighting a cross-lingual selection bottleneck that impedes effective multilingual LLM training.

Chenyu Zhou, Qiliang Jiang, Xu ZhouAug 24, 2026
AI Engineering & DevToolsAI Research

TriPLU Improves Tiny Language Model Performance with Trilinear Product FFNs

Researchers introduce TriPLU, a Trilinear Product Linear Unit, which replaces gated FFNs in tiny decoder-only language models with a direct degree-3 product branch. This approach achieves better validation loss on character-level TinyStories and lower bits per byte on other datasets under low-learning-rate settings, suggesting benefits for small models in specific low-compute regimes.

He ZhangAug 24, 2026