TheXpert: No. GA is external because it doesn't need to know the internals.
Yes it is external with respect to neurons, but if you screw it in as a learning method it must be internal with respect to the mesh, otherwise it will fall out of the overloaded learning methods.
Does genetics need to be able to change network topology? And how serious if they do?
This is the next step and is not directly related to the engine, its implementation lies through an external GA that creates different topologies, those initialize objects of different engines.
You can't make it shorter yet. You can get around this by creating different instances of the network and adding its description to the genes.
I was thinking about how to screw it in so that during training new neurons would be born in the right places, but after studying the algorithms I came to the conclusion that they will not have a common formalization, almost no points of intersection. That is why I refused, concentrating on the vertical construction of the network. By type:
Allow the use of cookies to log in to the MQL5.com website.
Please enable the necessary setting in your browser, otherwise you will not be able to log in.
You are signed in. Two-factor authentication has been disabled. Set it up again and scan a new QR code, or continue with two-factor authentication disabled.
No. GA is external because it doesn't need to know the internals.
Au genetics.
Does genetics need to be able to change network topology? And how serious if they do?
Au genetics.
Does genetics need to be able to change network topology? And how serious if they do?
No.
I was thinking about how to screw it in so that during training new neurons would be born in the right places, but after studying the algorithms I came to the conclusion that they will not have a common formalization, almost no points of intersection. That is why I refused, concentrating on the vertical construction of the network. By type:
data --> neuron --> encapsulated neural network --> container neural network
???
What's not to understand? Genetics can train any network, a network doesn't need to know about genetics at all.
doesn't have to, and won't, but GA will be part of the "container neural network" just like any other training method.
Just like a neural network doesn't have to be part of the GA, the GA only needs to know weights and error about the network, that's all.
I propose the first to implement either echo or SOM along the way and the interfaces can be finally settled.
Come on, what's closer to you?
The easiest thing to do is MLP