Current State and Future Directions for Learning in Biological Recurrent Neural Networks: A Perspective Piece
URL: “NA”
DOI: “NA”
tags: [Computer Science - Artificial Intelligence, Quantitative Biology - Neurons and Cognition] Read::
- Current State and Future Directions for Learning in Biological Recurrent Neural Networks: A Perspective Piece L.Y. Prince, R.H. Eyono, E. Boven, A. Ghosh, J. Pemberton, F. Scherr, C. Clopath, R.P. Costa, W. Maass, B.A. Richards, C. Savin, K.A. Wilmes 2022 🛫 2023-03-02 reading citation Print:: ❌ Zotero Link:: NA PDF:: NA Files:: arXiv.org Snapshot; Prince et al_2022_Current State and Future Directions for Learning in Biological Recurrent Neural.pdf Reading Note:: L.Y. Prince, R.H. Eyono, E. Boven, A. Ghosh, J. Pemberton, F. Scherr, C. Clopath, R.P. Costa, W. Maass, B.A. Richards, C. Savin, K.A. Wilmes (2022) Web Rip::
TABLE without id
file.link as "Related Files",
title as "Title",
type as "type"
FROM "" AND -"ZZ. planning"
WHERE citekey = "princeCurrentStateFuture2022"
SORT file.cday DESCAbstract
We provide a brief review of the common assumptions about biological learning with findings from experimental neuroscience and contrast them with the efficiency of gradient-based learning in recurrent neural networks. The key issues discussed in this review include: synaptic plasticity, neural circuits, theory-experiment divide, and objective functions. We conclude with recommendations for both theoretical and experimental neuroscientists when designing new studies that could help bring clarity to these issues.
Quick Reference
Top Comments
Let’s say grey is for overall comments
Tasks
Topics
Further Reading
—
Extracted Annotations and Comments
It’s well established that some connectivity motifs and microcircuits are conserved across species and individuals (e.g., neocortical columns, the hippocampal-entorhinal formation, cortico-cerebellar loops, and striatal networks). Furthermore, a certain degree of functional specialization consistently emerges across brain regions in terms of hierarchies within visual, somatosensory, motor, or auditory cortices. This degree of conservation indicates that much of the information required to generate these architectures are stored genetically, as a consequence of animal evolution. While panellists agreed that it is likely too extreme to say (in mammals at least) that learning relies on pre-wired recurrent circuits (as in echo state networks or liquid state machines [26]), there are clearly constraints imposed by the genetic code that may offer useful inductive biases for self-organization and learning. (p. 4)