On Fri, 2019-05-24 at 00:10 +0000, Alex via Digitalmars-d-announce wrote: > On Thursday, 23 May 2019 at 23:17:08 UTC, Murilo wrote: > > Guys I'm trying to make a program that simulates a neuron which > > behaves like the Physarum polycephalum so it will be able to > > develop intelligence. I'm making it totally in the D > > programming language. I will need help from biology nerds. If > > you want to participate you can contact me:
I am not an expert in this, but would a neuron (from whatever beastie) ever behave like a slime mould? > > 1- on my GitHub: https://github.com/MuriloMir > > or > > 2- via e-mail: [email protected] > > or > > 3- via the Dlang facebook group: > > https://www.facebook.com/groups/662119670846705/ > > or > > 4- on my twitter: https://twitter.com/MuriloMN0 > > It doesn't matter how you model a neuron. Whatever sigmoid > function you use will end up converging to the same result. All > neurons function in the same way, and that is as a switch. This > is why you can use all kinds of stuff for switches and it work. Is a sigmoid function sufficient? The era of treating a neuron as purely a single dimensional (electrical) state has, I believe, long past. Neurons do trigger, but they also have a biochemical aspect as well as an electrical one. I am not up to date with modelling neurons, and neither am I an expert in neurochemistry, and whilst investigating a network of sigmoid function triggers is still valid as a fun thing to do, I am not sure it can now be seen as a model of a collection of neurons. A model that started up in the mid to late 1970s but didn't take off then, but I believe is being picked up again recently, is to treat a network of neurons embedded in a biochemical system as a set of fields. The background was relativistic quantum field theory, but I suspect the technique as applied to networks of neurons has evolved away from that background. But maybe this is still not a mainstream approach? Does anyone have any connection with people working on Blue Brain. Over decade ago they were modelling the neocortex and neurons with apparently good success. > It seems that as long as they mimic a step function then it will > work. > > I'd suggest you design your algorithms around using a generic > neuron and then you can play around with specific implementations. -- Russel. =========================================== Dr Russel Winder t: +44 20 7585 2200 41 Buckmaster Road m: +44 7770 465 077 London SW11 1EN, UK w: www.russel.org.uk
signature.asc
Description: This is a digitally signed message part
