What is it about?
An efficient and robust fMRI decoding model with ingenious frequency domain design , which achieves an impressive 94.6% fMRI-to-image retrieval acc with 98.7% fewer parameters. We elaborately design a DFT backbone with Spectrum Compression and Frequency Projector modules to learn informative and robust voxel embeddings.
Featured Image
Photo by Growtika on Unsplash
Why is it important?
Single-subject lightweight and efficient model enables better application of decoding retrieval models.
Perspectives
hope that this paper will allow people to put some of their efforts on lightweighting and efficiency of brain decoding models. Current brain decoding models progress towards larger models, and we need to think about whether smaller models can also achieve comparable performance. I hope this paper can encourage researchers who do not have large GPU clusters to do so.
Zixuan Gong
Tongji University
Read the Original
This page is a summary of: Lite-Mind: Towards Efficient and Robust Brain Representation Learning, October 2024, ACM (Association for Computing Machinery),
DOI: 10.1145/3664647.3681229.
You can read the full text:
Resources
Contributors
The following have contributed to this page
