Sašo Džeroski

This paper introduces Kodiak, a discrete representation learning method for scientific imaging that improves representation quality and enables interpretable, reusable concepts across domains.

The paper compares canonical neural population models with grammar-based generated models to assess their ability to explain EEG data, finding that simpler models often perform better and that new models can match or exceed canonical ones.

The paper introduces a system to regulate protein function and create Boolean logic gates using coiled-coil dimers in protein loops, demonstrated on ten diverse proteins.

The paper reports results from the CAFA3 challenge, showing improved computational protein function prediction and new annotations for over 1000 genes through experimental screens in C. albicans, P. aureginosa, and Drosophila.