Richard A. Watson on Stress and Learning in neural networks and gene regulatory connections

you might say well gene networks get arranged by natural selection you change the networks at random and if they are fitter they get retained and if they're not they don't but neural networks change a different way neural networks change by learning a learning rule like neurons that fire together wire together changes the strengths of the connections between the neurons and that changes the behavior of the network Martin Anderson- So that's a different mechanism neurons that fire together wire together, whereas the way the gene networks change over time is just random variation selection.

Martin Anderson- But it turns out that actually the way in which variation and selection changes doing regulatory connections.

Martin Anderson- And.

experiences selective pressures which are the same as hepserol the same as neurons that fire together wire together so two genes which are expressed at the same time Dr Malcolm Johnston, M.D.: : which are selected to express at the same time, are also implicitly selected to increase the strength of their transcription factor connection between each other.

Dr Malcolm Johnston, M.D.: : Whereas if one gene is selected to be expressed when the other gene is selected to be off and vice versa, then a negative.

Dr Malcolm Johnston, M.D.: : connection between those two genes is favored by selection so selection changes gene networks in the same way that learning changes neural networks.

But when you think about, you know, both of those things, they seem sort of special, right? One involves natural selection, the other one involves a learning mechanism.

But it feels like, you know, that's interesting, but it's special.

It applies to neural networks, it applies maybe to gene networks, but it's not going to apply to other kinds of networks and other kinds of systems.