Nonlocal operator learning for fMRI encoding and decoding tasks
The paper discusses a novel approach to fMRI data analysis using nonlocal operator learning. It emphasizes the importance of spatiotemporal context in encoding and decoding tasks. The findings suggest that larger temporal windows enhance performance and representation learning in fMRI dynamics.
- ▪Functional MRI data presents challenges due to its high-dimensional spatiotemporal structure.
- ▪The study implements a latent neural integral operator framework for fMRI encoding and decoding tasks.
- ▪Results indicate that larger temporal windows generally improve classification and representation learning.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Machine Learning arXiv:2605.20389 (cs) [Submitted on 19 May 2026] Title:Nonlocal operator learning for fMRI encoding and decoding tasks Authors:Andreas Kramer, Saugat Acharya, Alice Giola, Emanuele Zappala View a PDF of the paper titled Nonlocal operator learning for fMRI encoding and decoding tasks, by Andreas Kramer and 2 other authors View PDF HTML (experimental) Abstract:Functional MRI data exhibit high-dimensional spatiotemporal structure, making both prediction and decoding challenging. In this work, we investigate neural integral-operator-based models for encoding and decoding tasks in fMRI, with particular emphasis on the role of nonlocal spatiotemporal context.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
