BrainTaskonomy optimizes fMRI foundation model pretraining and transfer learning

Researchers have developed a novel approach called BrainTaskonomy to optimize the pretraining and transfer learning processes for functional magnetic resonance imaging (fMRI) foundation models. This method utilizes a Brain-DiT proxy to estimate learning relationships between different fMRI domains, guiding the pretraining curriculum with a priority-guided approach and timestep scheduling. For downstream tasks, BrainTaskonomy constructs a directed taskonomy to select optimal source tasks and transfer routes using budgeted integer programming. This strategy significantly reduces various error metrics and demonstrates strong performance across multiple tasks, revealing asymmetric and target-dependent transfer patterns. AI

IMPACT This research could lead to more efficient and effective foundation models for analyzing brain imaging data, potentially accelerating neuroscience research.

RANK_REASON The cluster contains a research paper detailing a new methodology for fMRI foundation models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• Bip
• Brain-DiT
• BrainTaskonomy
• FC-MSE
• functional magnetic resonance imaging
• PSD-NMSE
• v-NMSE

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →