Learnijoy
ResearchAI ResearchAI Engineering & DevTools

MnemoDyn Model Learns Brain Dynamics from 40K fMRI Sequences

Sourav Pal, Viet Luong, Hoseok Lee, Tingting Dan, Guorong Wu, Richard Davidson, Won Hwa Kim, Vikas Singh· August 26, 2026 View original

Key takeaways

  • MnemoDyn is a compute-efficient, dynamical-systems based model for rs-fMRI.
  • It was trained on an unprecedented 40,000 fMRI sequences.
  • The model achieves superior reconstruction quality and generalizes well across diverse data.
  • It outperforms transformer-based approaches and is highly performant for downstream tasks.

Who benefits

HealthcarePharmaceuticalsResearch & AcademiaMedical Devices

Summary

MnemoDyn is a new dynamical-systems based model for resting-state fMRI, trained on a massive dataset of 40,000 sequences. It uses multi-resolution temporal modeling to achieve superior reconstruction quality and generalizes well across diverse populations and scanning protocols, outperforming transformer-based approaches.

Researchers have introduced MnemoDyn, a novel model designed to understand resting-state functional magnetic resonance imaging (rs-fMRI) dynamics. Unlike many contemporary approaches that rely on transformer architectures, MnemoDyn employs a dynamical-systems based framework with multi-resolution temporal modeling across parcellated brain regions. This model was trained on an exceptionally large dataset comprising approximately 40,000 rs-fMRI sequences, gathered from a wide array of public and permission-based datasets. The study highlights MnemoDyn's computational efficiency and its remarkable ability to generalize across diverse populations and varying scanning protocols. When benchmarked against current state-of-the-art transformer-based methods, MnemoDyn consistently demonstrates superior reconstruction quality. The findings suggest that large-scale pre-training on non-proprietary rs-fMRI datasets can yield highly performant models for various downstream neuroimaging tasks, particularly benefiting studies with small sample sizes where rs-fMRI is a common modality.

Why it matters

This model represents a significant step forward in analyzing brain activity from fMRI data, offering a more efficient and generalizable tool for neuroscience research and potentially for clinical applications. Its ability to perform well with small sample sizes is particularly valuable for rare conditions or specialized studies.

How to implement this in your domain

  1. 1Explore integrating MnemoDyn into neuroimaging research pipelines for fMRI data analysis.
  2. 2Evaluate MnemoDyn's performance on specific downstream tasks like biomarker discovery or neurological disorder prediction.
  3. 3Collaborate with research institutions to apply this model to existing fMRI datasets.
  4. 4Investigate the potential for MnemoDyn to enhance diagnostic tools in clinical settings.

Original post by Sourav Pal, Viet Luong, Hoseok Lee, Tingting Dan, Guorong Wu, Richard Davidson, Won Hwa Kim, Vikas Singh

"arXiv:2608.23936v1 Announce Type: new Abstract: We present a dynamical-systems based model for resting-state functional magnetic resonance imaging (rs-fMRI), trained on a dataset of roughly 40K rs-fMRI sequences covering a wide variety of public and available-by-permission datase…"

View on X

Originally posted by Sourav Pal, Viet Luong, Hoseok Lee, Tingting Dan, Guorong Wu, Richard Davidson, Won Hwa Kim, Vikas Singh 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 & DevToolsAI Investing

FraudBench Benchmarks Adversarial Robustness in Financial Risk Assessment

This paper introduces FraudBench, a protocol-sensitive benchmark for evaluating the adversarial robustness of machine learning models in financial fraud and credit-risk detection. It demonstrates that robustness conclusions are highly dependent on how domain-specific constraints and attacker capabilities are incorporated into the evaluation protocol.

Xitong Zeng, Zhaoge Bi, Yitian Yang, Huaming Chen, Quan Z. ShengAug 26, 2026
AI ResearchAI Engineering & DevTools

Persistent Cross Entropy Extends Topological Data Analysis

This paper introduces Persistent Cross Entropy (PCE), a novel extension of cross-entropy to persistence diagrams, which are used in topological data analysis. PCE bridges different event spaces of diagrams using an induced probability, enabling new applications like distinguishing diagrams with similar persistent entropy and separating causal directions in dynamical systems.

Sijin Yeom, Jae-Hun JungAug 26, 2026
AI ResearchAI Engineering & DevTools

Bridging Numerical PDE Solvers and Neural Emulators for Faster Simulation

This thesis explores the deep connections between traditional numerical solvers for Partial Differential Equations (PDEs) and neural emulators, arguing that they are more alike than different. It proposes that insights can flow profitably in both directions, leading to faster and more efficient scientific and engineering simulations.

Felix KoehlerAug 26, 2026