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

• DOI: 10.1073/pnas.2300558120
• Publication Date: 2023-07-31T19:07:46Z
• Authors: Willem A. M. Wybo; Matthias C. Tsai; Viet Anh Khoa Tran; Bernd Illing; Jakob Jordan; Abigail Morrison; Walter Senn
• Keywords: Hebbian theory; Feed forward; Representation; Learning rule; Modulation (music)

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