Abstract

Cerebellum-like structures -- such as the insect mushroom body -- are found in many brains and share a basic fan-out-fan-in network architecture. How the specific structural features of these networks give rise to their learning function remains largely unknown. To investigate this structure-function relationship, we developed a minimal computational model of the extensively studied Drosophila melanogaster mushroom body. We show how well-defined connectivity patterns between the Kenyon cells -- the constituent neurons of the mushroom body -- and their input projection neurons endow different functions, enabling the mushroom body to process olfactory information more efficiently. First, biases in the likelihoods at which individual projection neurons connect to Kenyon cells allow the mushroom body to prioritize the learning of particular, ethologically meaningful odors. Second, groups of projection neurons connecting preferentially to the same Kenyon cells facilitate the mushroom body to generalize across similar odors. Altogether, our results demonstrate how different connectivity patterns shape the representation space of a well-studied cerebellum-like network and impact its learning outcomes.