Talk:Synaptic pruning

Latest comment: 2 years ago by 2A02:AA1:1024:EBBE:10B4:B4E5:440E:8BBB in topic Remove section "Forgetting problems with learning through pruning"

Wiki Education Foundation-supported course assignment

edit

  This article was the subject of a Wiki Education Foundation-supported course assignment, between 10 September 2020 and 15 December 2020. Further details are available on the course page. Student editor(s): Fujeau.

Above undated message substituted from Template:Dashboard.wikiedu.org assignment by PrimeBOT (talk) 10:35, 17 January 2022 (UTC)Reply

Question

edit

Isn't pruning an analogy for these mechanisms rather than the name of them? — Preceding unsigned comment added by Hollth (talkcontribs) 15:16, 4 October 2014 (UTC)Reply

How about axonal pruning?

edit

How about synaptic pruning?

edit
  • Most of this article isn't about synaptic pruning but about neuronal pruning. I fully agree with the previous remark that it's probably better to rename this article and make three sections: synaptic pruning, axonal pruning, neuronal pruning. Each with references to seminal papers. —Preceding unsigned comment added by 130.37.96.116 (talk) 09:03, 8 April 2010 (UTC)Reply
  • I fully agree. The article promotes a very axon-centric view of cortical plasticity. Retraction of dendritic spines is not even mentioned as a synaptic pruning mechanism. Degeneration of entire axons is related to brain pathology, not to learning and fine-tuning of circuits. Needs a major rewrite.Millencolin (talk) 10:49, 12 January 2018 (UTC)Reply

Pregnancy

edit

The Washington Post published an article today[1] about how this process occurs in pregnant women, too. Interesting reading. 71.162.200.102 (talk) 03:55, 20 December 2016 (UTC)Reply

Planning to add section on Role in Psychiatric Disease

edit

I'm currently in the process of gathering sources to add a section on the role of synaptic pruning in schizophrenia and possibly ASD. I'm going to try to tie this in with a discussion of how maternal immune activation leading to overexpression of cytokines can result in increased synaptic pruning and possible contributions to these and related neurodevelopmental disorders. — Preceding unsigned comment added by Fujeau (talkcontribs) 17:02, 30 November 2020 (UTC) Fujeau (talk) 17:04, 30 November 2020 (UTC)Reply

Remove section "Forgetting problems with learning through pruning"

edit

Pruning in ANNs is essentially orthogonal to catastrophic forgetting. Catastrophic forgetting seemingly comes from the fact that ANNs use backprop to learn weights, where we don't have a written set of all local updating rules that the brain uses. Comparing primitive ANNs to the brain is premature, and the fact that ANNs are poor at pruning abilities almost surely says nothing of biology. — Preceding unsigned comment added by HedonistCalculus (talkcontribs) 22:33, 11 May 2022 (UTC)Reply

No, the problem has been repeated in multiple types of artificial neural networks with different mechanisms of learning weights that have nothing other in common than the fact that they rely on pruning. The problem is actually predicted by informational research on complex systems, from which it follows that it is the fact that the same synapses can, and to a greater extent in more complex systems more often do, form part of more than one possible signal pathway and more than one memory, that causes catastrophic interference. The effect is that in a more complex brain, memories become dependant on more junctions where a single junction pruned merely because it was somewhat rarely used or mistakenly pruned by a biological mechanism for pruning large numbers of synapses would permanently delete a much larger number of memories and abilities. A biological program that dooms as much as the 40 least used percent of the synapses to be pruned within a few years would not spare any brain of vertebrate complexity, let alone a human brain, from catastrophic interference since virtually all memories in complex brains pass theough so many synapses that even often recalled memories have some of the synapses they rely on among the overall least used synapses that happen to be less shared by other memories. This means that the problems with catastrophic interference would be worse, not milder, in complex biological brains than in simple artificial neural networks, provided similar reliance on pruning for development. Therefore, the section should not be removed. 2A02:AA1:1024:EBBE:10B4:B4E5:440E:8BBB (talk) 12:39, 17 June 2022 (UTC)Reply