Neural-Mass model acceleration on GroqChip [Theme: BrainFrame]

Mean-field approximation of network of biophysical neurons driven by  conductance-based ion exchange | bioRxiv

Epilepsy treatment is made possible using patient-specific virtual brains based on neural mass models. Reverse-engineering these models from MRI-scans and EEG-measurements requires substantial computational power. Currently, GPUs are used for this phase but don’t provide the required speed-up. We proudly house one of the few GroqChips available in the Netherlands. This chip (somewhere in-between an FPGA and a GPU) appears as an excellent choice for accelerating parameter explorations in parallel neural-mass models, crucial in full-brain modeling for epilepsy patients. While SpMV poses a bottleneck on GPUs, shifting SpMV instruction scheduling to the compilation phase, fitting it to the GroqChip paradigm, shows strong potential to significantly outperform current GPUs.

Job Category: MSc Thesis topic
Job Type: Student Thesis

Apply for this position

Allowed Type(s): .pdf, .doc, .docx