Beyond Single-Source Cognitive Taskonomy:Multi-Source Task Relations through fMRI Transfer Learning

2026-06-24

AI Summary

This study addresses the limitations of existing fMRI-based cognitive task classification approaches, which are largely confined to single-source transfer and fail to capture the complex influence of multi-source collaboration on target tasks. The work extends fMRI task classification to a multi-source transfer setting for the first time, leveraging masked fMRI reconstruction as a self-supervised objective to systematically investigate transfer relationships across 23 cognitive tasks. To optimize task selection under supervised budget constraints, the authors introduce Boolean Integer Programming (BIP). Experiments involving 1,127 models reveal strong within-domain transfer for motor tasks but weak cross-paradigm generalization, while working memory tasks—integrating perception, attention, and executive functions—are prioritized for supervision under resource constraints, underscoring their central role. The study further uncovers nonlinear effects of multi-source combinations on transfer performance and proposes a global-objective-driven strategy for supervised resource allocation.

Abstract

Cognitive tasks are organized by shared and specialized neural processes. Masked fMRI reconstruction provides a common self-supervised objective for quantifying transfer relations among task states, but existing reconstruction-based taskonomies mainly study one-to-one transfer from a single source task to a target. Here, we extend an fMRI cognitive taskonomy from single-source to multi-source transfer across 23 Human Connectome Project task states and use Boolean Integer Programming (BIP) to analyze budget-constrained task allocation. We train 1,127 task-specific and transfer models. Single-source transfer is directional and paradigm structured: motor states transfer well within the motor paradigm but provide limited support to most non-motor targets, consistent with a shared sensorimotor execution system and effector-specific representations. Multi-source transfer depends on the composition of the source set, suggesting that many-to-one task relations are not fully captured by pairwise taskonomy alone. Across supervision budgets, BIP repeatedly allocates direct supervision to several 0-back and 2-back working-memory states, although these states are not consistently the strongest individual sources. This pattern may reflect the integration of perceptual, attentional, and executive processes in working-memory tasks. Together, these findings reveal a cross-paradigm-limited motor cluster and working-memory states with high priority under the specified global allocation objective. Our study extends reconstruction-based fMRI taskonomy from one-to-one transfer to many-to-one task relations and budget-constrained task dependencies.

Problem

• fMRI transfer learning
• cognitive taskonomy
• multi-source transfer
• task relations
• supervision budget

Innovation

• multi-source transfer
• fMRI reconstruction
• taskonomy
• Boolean Integer Programming
• budget-constrained allocation