Super excited to announce Exploring fMRI RDMs: enhancing model robustness through neurobiological data has been published in the journal, Frontiers in Computer Science.
In the paper we preprocess the BOLD5000 (a large functional magnetic resonance imaging (fMRI) dataset) into representational dissimilarity matrices (RDMs), and establish an infrastructure researchers from a variety of backgrounds can use to train models with biologically grounded representations.
We then use this infrastructure ourselves to investigate the representations of several popular neural networks and apply a previously unexplored graph-based technique, Fiedler partitioning, to the BOLD5000 to test its biological viability.
The paper can be read here.
