Re:I should have loved biology
by MereInterest on Saturday July 09, 2022 @02:51PM
(#32038138)
There was a paper a few years ago about a similar effect in artificial neural networks [0]. The gist was that a large network can contain many subnetworks, and the number of subnetworks grows much faster than the size of the network they are contained in. They were able to find a subnetwork in a randomly weighted network with equivalent performance to a trained network of a much smaller size.
Nice. Sounds like these self-assembling subnets could be the basis for a viable model explaining the mechanisms behind early evolution.
Skynet?
Knowledge Distillation is a related concept in deep NNs, as are the concepts behind the compression of data in signal processing.
wow ... it makes sense ... more of a top down approach.
Its actually top down and bottom up at the same time. All of biochemistry operates on the basic rules of physics which determine how the chemistry happens with feedback from the surroundings/system as the top down part
Not sure if it was from a published paper, but the idea was that early proteins might have been large - say several hundred residues - but mostly disordered.
The smaller, more ordered 'domains' would then have evolved within these larger chains. Recombination and deletion would then have pruned down the disordered parts to leave more efficient smaller proteins.
No idea if that idea makes sense or has any research behind it, but it's quite a neat theory.