Institut Pasteur
Embodied neuroAI with drosophila larva
Animals are not free-floating neural networks that perform classification tasks. Our body, the temporal continuity of our actions, and our sensory word shape the property and the computing of biological neural networks. Advances in Drosophila genetic manipulation, neural connectomics, and amortised inference (acceleration of the sampling procedure using neural networks) have allowed us to study small neural circuits’ organisation, structure, and function. We still have a limited understanding of the forces organising small neural circuits.
In this presentation, I will address finite element simulations of the Drosophila larva body coupled to neural dynamics simulation of selected circuits, phenotyping larva as a statistical test in a learned latent space and neuromodulation at the single neuron scale.