Rhamba: Region-Aware Hybrid Attention–Mamba Framework for Self-Supervised Learning in Resting-State fMRI
Ruthwik Reddy Doodipala, Pankaj Pandey, Pratheek Eranki, Carolina Torres-Rojas, Manob Jyoti Saikia, Ranganatha Sitaram · 2026
Read Rhamba: Region-Aware Hybrid Attention–Mamba Framework for Self-Supervised Learning in Resting-State fMRI by Ruthwik Reddy Doodipala, Pankaj Pandey, Pratheek Eranki, Carolina Torres-Rojas, Manob Jyoti Saikia, Ranganatha Sitaram (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-01240.txt.
Description
Pankaj Pandey and five co-authors introduce Rhamba, a region-aware framework for self-supervised learning from resting-state brain scans. Pre-training is promising for large-scale neuroimaging, but region-aware masking and hybrid sequence modeling remain underexplored. Rhamba joins anatomically guided masking with hybrid attention and Mamba architectures, using region-aligned patch embeddings and three masking strategies of increasing spatial specificity.
Models are pre-trained on the ABIDE dataset and fine-tuned for schizophrenia and ADHD classification on other datasets. Masking strongly affects reconstruction loss in a consistent order, yet downstream differences are modest and dataset-dependent. A hybrid configuration with Mamba before attention achieves the highest average score. Rhamba beats leading methods overall. Peak performance depends on the interaction between masking and architecture.
Catalog record
| CREATOR | Ruthwik Reddy Doodipala, Pankaj Pandey, Pratheek Eranki, Carolina Torres-Rojas, Manob Jyoti Saikia, Ranganatha Sitaram |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 71 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.01240