Dendritic modulation enables multitask representation learning in hierarchical sensory processing pathways

Abstract

While sensory representations in the brain depend on context, it remains unclear how such modulations are implemented at the biophysical level, and how processing layers further in the hierarchy can extract useful features for each possible contextual state. Here, we first demonstrate that thin dendritic branches are well suited to implementing contextual modulation of feedforward processing. Such neuron-specific modulations exploit prior knowledge, encoded in stable feedforward weights, to achieve transfer learning across contexts. In a network of biophysically realistic neuron models with context-independent feedforward weights, we show that modulatory inputs to thin dendrites can solve linearly non-separable learning problems with a Hebbian, error-modulated learning rule. Finally, we demonstrate that local prediction of whether representations originate either from different inputs, or from different contextual modulations of the same input, results in representation learning of hierarchical feedforward weights across processing layers that accommodate a multitude of contexts.

Supralinear dendritic integration in murine dendrite-targeting interneurons

“…Accumulating evidence indicates that dendrites do more than merely summate excitatory and inhibitory synaptic inputs [1][2][3][4][5][6] . Dendrites endow neurons with increased computational capacity 7 and the ability to act as multi-level hierarchical networks 8,9 or even to reproduce some features of artificial and deep neural networks 10,11 . The underlying mechanisms include extensive dendritic arborisation, compartmentalisation, synaptic plasticity and expression of specific receptors and ion channels that in turn facilitate nonlinear input summation.…”

“…Nonlinear dendritic integration of excitatory inputs in particular can potentially support signal amplification 12 , coincidence detection 13 , XOR logic gating 14,15 , and computing prediction errors 2,11 . One striking example of non-linear integration is local supralinear summation of signals that arrive at a dendritic fragment in a narrow time window, a phenomenon that can be investigated by exploiting the spatial and temporal precision of somatic depolarizations elicited by multi-photon glutamate uncaging [16][17][18][19][20][21] .…”