Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster B in Poster Session B: Tuesday, August 4, 2:00 – 3:45 pm, Kimmel Center, Shorin & Rosenthal Rooms

Efficient Multi Subject Visual Reconstruction from fMRI Using Aligned Representations

Christos Zangos¹, Danish Ebadulla², Thomas Christopher Sprague¹, Ambuj Singh²; ¹University of California, Santa Barbara, ²UC Santa Barbara

Presenter: Christos Zangos

Reconstructing visual stimuli from fMRI remains a challenging problem, particularly under limited data and computational constraints, where existing methods degrade in performance. We introduce a subject-agnostic alignment framework using lightweight adapters, achieving up to 8% improvement in reconstruction metrics in low-data regimes. Combined with a principled image selection strategy, our approach reduces subject-specific data requirements from 1 hour to 24 minutes ( 60% reduction) while surpassing baseline performance and generalizing across architectures and training paradigms.

Topic Area: Methods, Tools, Theory & Neural Coding