Reference library
Dendritic cortical microcircuits approximate the backpropagation algorithm
Paper summary
A compartmental cortical-circuit model represents apical error as top-down feedback minus predictive interneuron inhibition. A self-predicting state cancels expected feedback, and output nudging unmasks an instructive residual. Approximation to backpropagation requires suitable small-feedback and predictive-state assumptions; symmetric, random and learned feedback cases differ. Continuous-time regression avoids mandatory phases, while classification demonstrations use computational shortcuts.
Relevance to the project
Useful for specifying what an instructive feedback channel actually computes and for distinguishing a local update from an exact downstream gradient. It gives a comparison for future gradient alternatives while leaving the adopted scorer preparation and U_F/U_H handover intact.
Computational evidence, not measured synaptic learning in an animal. The classification experiment's bottom-up initialization, reverse-order compartment updates and minibatches should not be treated as a wholly continuous biological demonstration. All proofs and supplements were not audited.
Source record
- Reference ID
- neurips-sacramento2018
- Archived version
- Published NeurIPS 2018 conference paper, proceedings PDF, 12 pages.
- Summary provenance
- reference-library/papers/neurips-sacramento2018/metadata.json · summary
- Review status
- Targeted source-grounded review; no experimental reproduction or exhaustive proof/figure audit
- Reading scope
- Published conference PDF acquired. Selected model/methods, regression/classification results, feedback controls and conclusions, principally pages 1–9; no complete derivation, supplementary-material or independent implementation audit.