You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
So there can be, somewhere, feedbacks?
The delay operator cells are responsible for the feedback; these are the convoluted out cells taken from the out array itself.
The only inconvenience is that for each rank of delay operator you have to allocate memory for the whole matrix, but it's a matter of optimization.
It will be possible to make loading templates by names, for example the template "MLP" prescribes the types of all neurons "perseptrons".
Well, if the idea is clear, then I repeat the main question for what I started it all:
Opponents are invited to present the type of grid that is not possible to describe this model!!!
it is necessary to understand whether the idea has a future, and then we will write a bunch of code, and then we will find out that it is impossible to implement many variants of grids.
Opponents are invited to present the type of grid that is not possible to describe with this model!!!
to abstract away from that very "grid type" - a rather artificial concept, if you get it right.
That's a hell of a way to put it... And synapses to grow and cut off. And billions of neurons. And the forest chops and stacks itself.
(Well, uh-huh.))
I wrote to you. Why do not you respond?
First, a little off-topic. There was such a project in Caltech, PolyWorld. The goal of the project was to create a two-dimensional world of virtual creatures, give them a simple brain-network, and the ability to see, move, find food, reproduce, kill each other (the amount of food is limited), etc. in order to survive. The brain-network consisted of several neurons with a given input-output function and plastic (learnable) synapses between neurons. At birth, the connections between neurons are random. The goal of the project was to observe the evolution of this virtual world. Those interested can watch a lecture about this project here (the interesting part starts at minute 16):
https://www.youtube.com/watch?v=_m97_kL4ox0
Now on to the subject. Has anyone tried to apply genetics to a network of neurons connected to each other, to inputs and outputs via initial-random weights? Synapses can appear and disappear by applying Hebb's rule or other unsupervised learning rule. Create a virtual world of such networks-traders trading forex, each having a certain initial deposit and trying to increase it at the expense of other traders. For example, every time when some trader makes a profitable deal, the profit flows from the funds of other traders, or something like that, but the total funds remain constant. Traders compete with each other for survival, i.e. not to drain the deposit. You can introduce the possibility of crossing network traders, etc. In short, create a real environment for the evolution of networks in the market.
I think I answered everything, maybe I missed something, if you don't mind repeating it, if I didn't answer.