Dendritic cortical microcircuits approximate the backpropagation algorithm

  • 2018-10-26 15:40:58
  • João Sacramento, Rui Ponte Costa, Yoshua Bengio, Walter Senn
  • 26


Deep learning has seen remarkable developments over the last years, many ofthem inspired by neuroscience. However, the main learning mechanism behindthese advances - error backpropagation - appears to be at odds withneurobiology. Here, we introduce a multilayer neuronal network model withsimplified dendritic compartments in which error-driven synaptic plasticityadapts the network towards a global desired output. In contrast to previouswork our model does not require separate phases and synaptic learning is drivenby local dendritic prediction errors continuously in time. Such errorsoriginate at apical dendrites and occur due to a mismatch between predictiveinput from lateral interneurons and activity from actual top-down feedback.Through the use of simple dendritic compartments and different cell-types ourmodel can represent both error and normal activity within a pyramidal neuron.We demonstrate the learning capabilities of the model in regression andclassification tasks, and show analytically that it approximates the errorbackpropagation algorithm. Moreover, our framework is consistent with recentobservations of learning between brain areas and the architecture of corticalmicrocircuits. Overall, we introduce a novel view of learning on dendriticcortical circuits and on how the brain may solve the long-standing synapticcredit assignment problem.


Introduction (beta)



Conclusion (beta)