Discussion of article "Third Generation Neural Networks: Deep Networks" - page 11
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
Meta Trader 4 tester generates an error:
i_SAE EURUSD,M30: array out of range in 'i_SAE.mq4' (140,22)
Rterm starts correctly, two instances.
According to the code it is Time[]. I made the indicator run by bars, by ticks, initialised together with the Expert Advisor. It doesn't help.
Is there a solution?
As a rule, it happens when the history to the left of the trigger point is less than "back" (i.e. less than 700 bars).
You cannot decrease this value. This is the minimum value.
Such errors have not popped up for a long time.
Check or pump up the history. And do not run the indicator at the very edge of the history.
Good luck
Nothing special, just initialised Expert Advisor and indicator in the console and in studio (R project, all according to the instructions in the article).
Is it possible to get mt4Rb7.dll for 64 bit and mql5?
It doesn't work without it, only in metatrader 4.
I haven't managed to add a tester yet (mql4).
R version 3.2.2
MT5 was not involved and do not plan to. DLL now works with 64-bit R in MT4 (32p).
According to my understanding, DLL work in MT5 should not cause any problems. Although I don't know the peculiarities of MT5.
And in what do you see the advantage of using MT5? What is impossible to do in MT4 compared to MT5? So for general development.
Good luck
Errors seem to be gone. But the second instance of Rterm is not launched when calling the indicator from the Expert Advisor via iCustom(). Is there any other way to launch the indicator together with the Expert Advisor, except including the indicator in the Expert Advisor?
About MQL5. I just started to master the language from it.
From what I have learnt. There is an undoubted advantage of creating an Expert Advisor based on indicator voting (signals), exchange of indicator buffers.
Of course, everything is implemented in MQL4, but through "crutches", not out of the box. There are also some difficulties when transferring code from MQL4 to MQL5 because of the lack of basic simple functions in MQL5, which were needed earlier, but are now hidden away.
Regarding mt4Rb7.dll. This library is compiled as a 32 bit system object and therefore will not run in Meta Trader 5, as it works as a 64 bit library. Only the 32-bit version of Meta Trader is suitable, the fifth version of which I have not found. That is why I am testing in the 4th version.
Errors seem to be gone. But the second instance of Rterm is not launched when calling the indicator from the Expert Advisor via iCustom(). Is there any other way to launch the indicator together with the Expert Advisor, except including the indicator in the Expert Advisor?
About MQL5. I just started to master the language from it.
From what I have learnt. There is an undoubted advantage of creating an Expert Advisor based on indicator voting (signals), exchange of indicator buffers.
Of course, everything is implemented in MQL4, but through "crutches", not out of the box. There are also some difficulties when transferring code from MQL4 to MQL5 because of the lack of basic simple functions in MQL5, which were needed earlier, but are now hidden away.
Regarding mt4Rb7.dll. This library is compiled as a 32 bit system object and therefore will not run in Meta Trader 5, as it works as a 64 bit library. Only the 32-bit version of Meta Trader is suitable, the fifth version of which I have not found. That's why I am testing in the 4th one.
Allow me my five cents.
Why is it so difficult with R? Some indicators, servers...
After all, everything works (at least for me) through the included library. You write a string variable in µl, which is a function call, pass it to R, which executes the necessary number of R functions, and happiness has come....
Or am I missing something?
Allow me my five cents.
Why is it so difficult with R? Some indicators, servers...
After all, everything works (at least for me) through the supplied library. You write a string variable in µl, which is a function call, pass it to R, which executes the necessary number of R functions, and happiness has come....
Or am I missing something?
Apparently a misunderstanding.
I want to run all this in the mql4 tester.
Without the tester, everything works well, happiness is there.
In MQL5 it does not work, because of the different digitisation of the objects to be run.
What is the difference between the attached library for working with R from the article and the one you suggest (weight is different)?
Apparently there is a misunderstanding.
I want to run in mql4 tester all this stuff.
Without the tester, everything works well, happiness is there.
In MQL5 it does not work, because of the different digitisation of the objects to be launched.
What is the difference between the attached library for working with R from the article and the one you suggest (weight is different)?
I am using this library now.
On the previous version of MT4 (up to 540) I used the tester without any problems.
The library itself is written in pascal, source code is available. I don't see any problems using this library in mcl4 or mcl5.
I think the whole problem is in the intricate circuitry where the tester shuts down
Allow me my five cents.
Why is it so difficult with R? Some indicators, servers...
After all, everything works (at least for me) through the supplied library. You write a string variable in µl, which is a function call, pass it to R, which executes the necessary number of R functions, and happiness has come....
Or am I missing something?
Greetings SanSanych.
The client-server variant is attractive for several reasons:
- working on many tools at the same time.
- But the main thing is that the server allows two-way non-blocking exchange of information not only client<->server (clients can be up to 128) but also between clients and other servers. I.e. during long calculations data exchange with the server and other clients is available. The only limitation is that you cannot request the result of a calculation before it is finished, Rterm will crash.This allows you to build a complex hierarchical system with one Rterm controlling all the others. Of course if you need it.
- It is possible to pass the data to an indicator, which will draw the necessary information on the chart. My experience says that there is no more reliable control than visual control. And drawing from the Expert Advisor is problematic.
Of course, everything can be collected in the Expert Advisor, but when working on several pairs, some problems arise, which of course can be solved in other ways.
This is one way of working, but it doesn't mean that it is the only or ideal one.
I try to show the maximum number of possible variants of work in the article. And the user will choose what he needs.
From the complex is easy to make simple, the opposite is difficult. Although in my opinion the proposed client-server variant is not so complicated. Look at the amount of code on MKL5 that is needed to solve such a problem.
Good luck
Apparently there is a misunderstanding.
I want to run in mql4 tester all this stuff.
Without the tester, everything works well, happiness is there.
In MQL5 it does not work, because of the different digitisation of the objects to be launched.
What is the difference between the attached library for working with R from the article and the one you offer (weight is different)?
Errors seem to be gone. But the second instance of Rterm is not launched when calling the indicator from the Expert Advisor via iCustom(). Is there any other way to launch the indicator together with the Expert Advisor except including the indicator in the Expert Advisor?
What indicator are we talking about? What indicator do you want to use via iCustom()?
Regarding MQL5. I just started to master the language from it.
From what I have learnt. There is an undoubted advantage of creating an Expert Advisor based on indicator voting (signals), exchanging indicator buffers.
All calculations, including those of indicators, should be done in R. And there are no restrictions for your fantasies in terms of mathematics. Only quotes are taken from MT!
Of course, everything can be implemented in MQL4, but through "crutches", not out of the box. There are also some difficulties when transferring the code from MQL4 to MQL5 due to the lack of basic simple functions in MQL5, which were needed earlier, but are now hidden away.
Regarding mt4Rb7.dll. This library is compiled as a 32 bit system object and therefore will not run in Meta Trader 5, as it works as a 64 bit library. Only the 32-bit version of Meta Trader is suitable, the fifth version of which I have not found. That's why I'm testing in the 4th one for now.
Try to write to the author. He has a branch where he appears from time to time. But according to my information, the library has not been redesigned for MT5.
You need to rearrange your ideas about the Expert Advisor's structure a bit. An Expert Advisor should do its job (execute orders, monitor positions, etc. etc.). To realise these tasks, it doesn't matter what language (MKL4 or MKL5) you will implement it in. It is a matter of preference. Calculations, analyses and other complex tasks need to be implemented in R process.
Separate "doing" from "thinking". No MKL is designed to "think". It is honed for "doing."
You need to use each language for what it was designed for.
Good luck
Vladimir Perervenko
Add additional information on working with R Studio to the article