NMDA-driven dendritic modulation enables multitask representation learning in hierarchical sensory processing pathways

While sensory representations in the brain depend on context, it remains unclearhow such modulations are implemented at the biophysical level, and how processinglayers further in the hierarchy can extract useful features for each possible contex-tual state. Here, we demonstrate that dendritic N-Methyl-D-Aspartate spikes can,within physiological constraints, implement contextual modulation of feedforwardprocessing. Such neuron-specific modulations exploit prior knowledge, encoded instable feedforward weights, to achieve transfer learning across contexts. In a network ofbiophysically realistic neuron models with context-independent feedforward weights,we show that modulatory inputs to dendritic branches can solve linearly nonseparablelearning problems with a Hebbian, error-modulated learning rule. We also demonstratethat local prediction of whether representations originate either from different inputs,or from different contextual modulations of the same input, results in representationlearning of hierarchical feedforward weights across processing layers that accommodatea multitude of contexts