Role of complex synaptic dynamics for learning and memory. (360G-Wellcome-110124_Z_15_Z)

How can real brains learn to perform diverse tasks so much better than supercomputers? One of the reasons is that synapses, the brain's storage medium for experience and memories, are too simplistic in our models. Commonly described only by single numbers denoting their strength, we have ignored the rich internal dynamics of real synapses. By harnessing this complexity, biological systems outperform today's most powerful algorithms. Here, I am proposing a paradigm shift towards complex synaptic models to study their dynamic role for the emergence of computation and memory in neural networks. AIM1: I will design and collaborate on experiments to test fundamental assumptions and predictions from the few existing and adverse theoretical works on complex synapse models. AIM2: Building on these findings, I will create a unifying theoretical framework for complex synapses and fit plausible synapse models to experimental data. AIM3: Finally, I will build biologically inspired spiking neural network models with complex synapses to study self-organization, learning and memory. Using these models, I will generate system-level predictions for experimental verification. Specifically I will collaborate on experiments in flies and mice, suitable for genetic perturbation studies; later on I may include human fMRI studies because of their medical relevance.