Advisors on neural networks, sharing experiences. - page 7

[Sergey Ponomarev]

Anton Govorukhin:
Yesterday I ran the neuroner on the demo https://www.mql5.com/ru/signals/129790 It doesn't seem to be doing anything yet.

Really, 'cause the signal's off.

[[Deleted]]

Sergey Ponomarev:
what is it

Advisors on neural networks?

It is the same as the usual ones, only the signals are not produced by a pre-computed algorithm, but by a trained neural network, like a black box :)

[Evgeniy Sergeev]

A neural network in the sense is an approximation of a table-defined function. And Forex price is a random value, we may as well train a neural network by inputting a random number. Although in the case of a random number the neural network will "calculate" the generation algorithm, but in Forex it will not work.

[Alexey Burnakov]

Evgeniy Sergeev:

A neural network in the sense is an approximation of a table-defined function. And Forex price is a random value, we may as well train a neural network by inputting a random number. Although in the case of a random number the neural network will "calculate" the generation algorithm, but in Forex it will not work.

In Forex the neuronet will calculate "an invisible hand of the market". )

[[Deleted]]

Evgeniy Sergeev:

A neural network in the sense is an approximation of a table-defined function. And Forex price is a random value, we may as well train a neural network by inputting a random number. Although in the case of a random number the neural network will "calculate" the generation algorithm, but in Forex it will not work.

No well, "random" will not work, it depends on something, so it is not random. And this something, in its turn, depends on something else.

[Sergey Chalyshev]

All are dummies. Can anyone show what your neural network shows during the training period?

[TheXpert]

Sergey Chalyshev:
All are dummies. Can someone show me what your neural network shows in the training period?

On the learning curve? Not even gold -- diamond mountains.

[TheXpert]

Yousufkhodja Sultonov:

Yusuf, with all due respect, go with your theory in your own thread.

[Sergey Chalyshev]

Комбинатор:
On the training period? Not even gold -- diamond mountains.

Show me then, very interesting. Preferably with details, number of neurons, number of inputs, number of training examples, etc.

[Evgeniy Sergeev]

Maxim Dmitrievsky:
Нет ну "случайная" - так не пойдет, от чего-то зависит значит уже не случайная. А это что-то, в свою очередь, от чего-то другого.

Then it's chaotic.

Have you tried to solve the problem in a different direction - not to predict the next bar, but to count the number of network errors? That is, the grid finds some regularity (trend) in the recent history and checks if this regularity is still present on the current bars. If the regularity stopped working (the net began to make too many errors), then the trend must have changed. And therefore we can look for an entry point into the market.