Abstract
In the last decade dendrites of cortical neurons have been shown to nonlinearly combine synaptic inputs by evoking local dendritic spikes. It has been suggested that these nonlinearities raise the computational power of a single neuron, making it comparable to a 2-layer network of point neurons. But how these nonlinearities can be incorporated into the synaptic plasticity to optimally support learning remains unclear. We present a theoretically derived synaptic plasticity rule for supervised and reinforcement learning that depends on the timing of the presynaptic, the dendritic and the postsynaptic spikes. For supervised learning, the rule can be seen as a biological version of the classical error-backpropagation algorithm applied to the dendritic case. When modulated by a delayed reward signal, the same plasticity is shown to maximize the expected reward in reinforcement learning for various coding scenarios. Our framework makes specific experimental predictions and highlights the unique advantage of active dendrites for implementing powerful synaptic plasticity rules that have access to downstream information via backpropagation of action potentials.
Highlights
One of the fascinating and still enigmatic aspects of cortical organization is the widespread dendritic arborization of neurons
Error-backpropagation is a successful algorithm for supervised learning in neural networks. Whether and how this technical algorithm is implemented in cortical structures, remains elusive. We show that this algorithm may be implemented within a single neuron equipped with nonlinear dendritic processing
An error expressed as mismatch between somatic firing and membrane potential may be backpropagated to the active dendritic branches where it modulates synaptic plasticity. This changes the classical view that learning in the brain is realized by rewiring simple processing units as formalized by the neural network theory
Summary
One of the fascinating and still enigmatic aspects of cortical organization is the widespread dendritic arborization of neurons These dendrites have been shown to generate dendritic spikes [1,2,3] that support local dendritic processing [4,5,6,7], but the nature of this computation remains elusive. We show that the dendritic morphology offers a substantial additional benefit over the 2-layer network This is because it allows for the implementation of powerful learning algorithms that rely on the backpropagation of the somatic information along the dendrite that, in a network of point neurons, would not be possible in this form. Transmit information just in one direction, making it difficult to implement error-backpropagation in biological neuronal circuitries In the 2-layer structure of a dendritic tree information at the output site may be physically backpropagated across the intermediate computational layer to the synapses targeting the tree
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
