ATcl - Tcl interpreter for MT4 - page 5
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
ATcl Beta 2 release
With a slight delay, a new version of ATcl has been released.
Release details, see http://nektomk.ru/atcl:beta2
Since there were problems with the domain, to make the distribution available, a project on SourceForge has been made: https://sourceforge.net/projects/mt-atcl/
The project can be downloaded from the SF page or from the usual download page: http://nektomk.ru/atcl:install
I am also attaching the archive
New features:
- WebSocket handling demo added
- Excel management and data exchange demo added
- Tk is now up and running (demo is also available) - now you can quickly and easily make the most sprawling GUI
Known bug:
- At the last moment was detected: there is a conflict with simultaneous loading of ATcl indicators and EA using Tk. The nature of the error is clear, but it has not been logged in the code yet and therefore it has not been fixed yet.
Separately, they all work fine, so I've decided to release it with this bug.
PS. Whether site glitches or browser, but this message I fill 3rd time.
And how to rebuild the assembly for x64? Or x64 is originally not supported?
How do I rebuild for x64? Or x64 is initially not supported?
for 64 (i.e. for MT5) there is no version yet. In the long term we have one, but not yet :-)
I wrote somewhere above - it's much harder to debug DLL for 5 this one, and I don't consider 5 to be stable and popular yet that's two.
And finally and most importantly - before you port to 64 bits the current API has to be completely stabile and acceptable.
Maintaining, developing and debugging 1 version is one thing, developing 2 is a bit harder, to put it mildly.
for 64 (i.e. for MT5) there is no version yet. In the long term we have it, but not yet :-)
I've written above somewhere - it's much harder to debug DLLs for 5 this one, and I don't think 5 is sustainable and popular yet that's two.
And finally and most importantly - before you port to 64 bits the current API has to be completely stabile and comfortable.
Maintaining, developing and debugging 1 version is one thing, while 2 versions is a bit more complicated, to put it mildly.
I see.
But to whom neural networks ? :-)
It's pretty easy to add to ATcl (so in MT4) the FANN - Fast Artificial Neural Network Library - http://leenissen.dk/fann/wp/.
Actually just attach a DLL and check its API
If anyone is interested, we can discuss and test-case will be needed .
It would be possible to run training and NN in EAs/indicators in the current track or in a separate one (or even many) in the background.
But for whom neural networks ? :-)
It's pretty easy to add to ATcl (so in MT4) FANN - Fast Artificial Neural Network Library - http://leenissen.dk/fann/wp/.
Actually just attach a DLL and check its API
If anyone is interested, we can discuss and test-case will be needed .
It would be possible to run training and NN work in EAs/indicators in the current track or in a separate one (or even many) in the background.
Sounds interesting but not clear :)
Sounds interesting, but not clear :)
point by point
- There is a FANN library for all sorts of NN. It's quite popular and they even tried to use it in MT (there's an article about it). But traditionally it was abandoned and buried in the annals of CodeBase.
- I've made a direct interface MQL4 <-> Tcl. They work in a shared address space and know how to exchange and share data. Tcl is a very-very old, well-known and stable language. Therefore many tools (especially FANN) have an API with it.
At the very least there are semi-automatic build/connect tools just by .h (C/C++ header files)
- ATcl development has reached a happy moment when multi-thread can be used, i.e. threads can be created and managed to run in background/parallel to main task, e.g. an Expert Advisor
- Direct use of threads - long or busy calculations. When there are a lot of calculations to be performed, but the chart or the terminal cannot be suspended. Neural networks are popular and a good candidate for trying this out.
- I'm willing to include FANN in ATcl, i.e. compile/create/debug interfaces and test in a multi-tasking environment. I'm interested in this in the practice of connecting libraries and debugging-debugging a multitasking environment. I don't use NN by itself in trade,
- if anyone has an interest, let's do it.
item by item.
- There is a FANN library for all sorts of NN. It's quite popular and has even been tried in MT (there's an article about it). But traditionally it was abandoned and buried in the annals of CodeBase.
- I have made a direct interface MQL4 <-> Tcl. They work in a common address space and know how to exchange and share data. Tcl is a very-very old and stable language. Therefore many tools (especially FANN) have an API with it.
At the very least there are semi-automatic build/connect tools just by .h (C/C++ header files)
- ATcl development has reached a happy moment when multi-thread can be used, i.e. threads can be created and managed to work in background/parallel to main task, e.g. an Expert Advisor
- Direct use of threads - long or busy calculations. When there are a lot of calculations to be performed, but the chart or the terminal cannot be suspended. Neural networks are popular and a good candidate to try this.
- I'm willing to include FANN in ATcl, i.e. compile/create/debug interfaces and test in a multi-tasking environment. I'm interested in this in the practice of connecting libraries and debugging-debugging a multitasking environment. I don't use NN by itself in trade,
- if anyone has an interest, let's do it.
But I am a simple user - I don't understand anything so far, only that it is possible to connect a neuron, but what is the advantage between the method described in the article and yours?
I'm interested in introducing neuronics in MT5, is it possible?
I'm a simple user - so far I don't understand much, only that it is possible to connect a neuron, but what is the advantage between the method described in the article and yours?
I am wondering if it is possible to implement neuronics in MT5?
The perceived benefits are that
1) the "integrator" would be tcl, which is very stable (and mildly older than MQ). And FANN and tcl have a mutual API which is developed not only by the mql5.com community
2) training and running of the neuron can be done in a separate thread, i.e. do not "hold" Expert Advisor/indicator
3) I see a smooth transition to MT5 - only 1 (one!) thing is subject to porting, i.e. ATcl interface, while all users' scripts and programs can be unchanged and directly shared/shared between Mt4-Mt5
the perceived advantages are that
1) the "integrator" will be tcl, which is very stable (and older than MQ, to put it mildly). And FANN and tcl have a mutual API, which is developed not only by the mql5.com community.
2) training and running neuronics can be done in a separate thread, i.e. not "delayed" EA/indicator.
3) I see a smooth transition to MT5 - only 1 (one!) thing is ported, i.e. ATcl interface, while all users scripts and programs can be unchanged and directly shared/shared between Mt4-Mt5
This is good, but now the main question - is it possible to optimize strategies of variables (predictors) in the tester, when coupled with neuronics?
This is good, but now the main question - is it possible to optimise strategies of variables (predictors) in the tester when coupled with neuronics?
In terms of software technology, potentially yes. Libraries can be called from the optimizer as long as it is not in the cloud.
from the NS side - how do you imagine it?