Prediction error drives associative olfactory learning and conditioned behavior in a spiking model of Drosophila larva

Abstract Predicting reinforcement from the presence of environmental clues is an essential component of guiding goal-directed behavior. In insect brains, the mushroom body is central to learning the necessary associations between sensory signals and reinforcement. We propose a biologically realistic spiking network model of the Drosophila larva olfactory pathway for the association of odors and reinforcement to bias behavior towards approach or avoidance. We demonstrate that prediction error coding through the integration of currently present and expected reinforcement in dopaminergic neurons can serve as a driving force in learning that can, combined with a synaptic homeostasis mechanism, account for experimentally observed features of acquisition and loss of associations in the larva that depend on the intensity of odor and reinforcement and temporal features of their pairing. To allow direct comparisons of our simulations with behavioral data [1], we model learning-induced plasticity over the complete time course of behavioral experiments and simulate the locomotion of individual larvae towards or away from odor sources in a virtual environment.