A quick and free library for MT4, much to the delight of neuralnetworkers - page 10
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
Removed the commented.
Now I explain, at res>0, here 0 corresponds to level 50 of the RSI indicator, we assign 1, otherwise we assign -1
What is not clear? A minimum of code and nothing else.
As it is written in the source code, the main condition of data normalization is NOT fulfilled.
The only thing is to filter out another zero, but in this case it's not that important.
In this case, why do you need the network? You teach it to trade following the algorithm of breakdown of zero line of normalized data (or breakdown of RSI of level 50). That is, you know the algorithm a priori - so program it and do not play with the network. Another thing is if you do not know the algorithm, and you try to make the network to find it, using the results of previous trades and relevant indicator readings.
Good luck.
Thanks, I see, the key word in all this is range.
You should probably sleep at night...
Normalised values (1;-1) must be fed to the NS input. Otherwise, training the NS can lead to uncertain results.
I would make a less bold statement - limited values should be fed to the NS inputs.
Good afternoon all, the optimization graph is not drawn, after optimization the following line appears
2009.12.21 15:52:54 There were 897 passes done during optimization, 897 results have been discarded as insignificant
can anyone help?
Good afternoon all, the optimisation graph is not drawn, after optimisation a string is displayed
2009.12.21 15:52:54 There were 897 passes done during optimization, 897 results have been discarded as insignificant
can someone help?
Try in the properties of the espert, on the Optimization tab, to remove all restrictions on the optimization results.
Просветите плиз, зачем умножение на 2 в строке 190:
You can comment this line altogether. It has no meaning. It was left by previous Expert Advisor.
Doesn't it? The string fills in the value returned by the ann_pnn function, and depending on it opens a buy or sell. Following that logic, the whole ann_pnn function is unnecessary, and let the orders open randomly. I also don't quite understand why grids are only trained on losing options (if (OrderProfit() < 0)).
I will try to dig even deeper. My log shows that the answers of all the grids are the same in one survey and different in the other, but the same. It's the same all the way through the control test.
14:44:45 2008.02.01 00:05 FANN-EA USDJPY,M5: f2M_get_output(0) returned: 0.05168430
14:44:45 2008.02.01 00:05 FANN-EA USDJPY,M5: f2M_get_output(1) returned: 0.05168430
........
14:44:45 2008.02.01 00:05 FANN-EA USDJPY,M5: f2M_get_output(14) returned: 0.05168430
14:44:45 2008.02.01 00:05 FANN-EA USDJPY,M5: f2M_get_output(15) returned: 0.05168430
I'll look into it, but who knows? What is wrong?
Re-training is likely. This advisor should not be seen as a guide to action - if one does understand it, it is more about not doing what it says. In particular, you should never follow the advice to use the genetic optimiser (as written on the main page https://www.mql5.com/ru/code/9386). It should only be used to optimize the weights of the grid itself (as was done in the perceptron example posted long ago on the site), and in case of selection of input parameters (which is done in the current FANN-EA), one should provide examples that are as uniformly distributed in the feature space as possible. If you include genetics, the grid will only chisel with the best examples.
Basically, the topic of neural networks interests many traders, but few realize that it cannot be dealt with lightly ;-). And some articles are written here, but either they are not enough, or no one really understands them.
Try in the properties of the espert, on the Optimisation tab, to remove all restrictions on the optimisation results.
All the checkboxes are unchecked there, what else could be the problem ? thanks for the reply :)
Попробую копнуть еще глубже. У меня по логу видно что ответы всех сеток одинаковы при одном опросе, при другом -- другие, но то же одинаковые. И так на всем протяжении контрольного теста
14:44:45 2008.02.01 00:05 FANN-EA USDJPY,M5: f2M_get_output(0) returned: 0.05168430
14:44:45 2008.02.01 00:05 FANN-EA USDJPY,M5: f2M_get_output(1) returned: 0.05168430
........
14:44:45 2008.02.01 00:05 FANN-EA USDJPY,M5: f2M_get_output(14) returned: 0.05168430
14:44:45 2008.02.01 00:05 FANN-EA USDJPY,M5: f2M_get_output(15) returned: 0.05168430
Буду разбираться, но может кто в курсе? Что не так?
In this EA, all of the committee networks are given the same input signal and require the same response. It is not surprising that the nets converge to the same solution. In this example you can leave one grid or modify the input system so that different nets are fed different inputs, the outputs can be left the same.
Good luck.