Martin Schrimpf receives prestigious SNSF SPARK grant

The project, entitled "Scaling Laws for Neural Network Models of the Primate Visual Ventral Stream", aims to revolutionize the modeling of the brain using artificial neural networks. While current trial-and-error approaches to improving brain models are computationally intensive and lack clear direction, this research proposes a novel strategy: neuroscientific scaling. Inspired by advances in machine intelligence, the project will establish scaling laws to predict how model performance aligns with brain activity as the size of models, datasets, and computational power increases.

By focusing on improving the alignment of neural networks with the primate visual cortex—rather than just optimizing performance on computer vision tasks—this approach promises a more efficient development of brain models. These scaling laws will enable researchers to test models on a smaller scale before expanding only the most promising configurations. The outcomes could deepen our understanding of the brain and mind while paving the way for transformative clinical applications.