AI "Artificial Intelligent" EA

 

How do you design a learning EA that will not consume alot of memory and diskspace with MQL4?


I have designed a learning "AI" EA before; believe me, they yield tremendous profits.

They can learn and distinguish the difference between good entries and bad entries way better than any human!


I have backtested my AI on several different pairs just to make sure I was not cherry picking.

I also tried different years. And my, the results were astounding and way into the sky!


My only problem is that the EA hogs lots of memory and is real slow.

Any idea on how to speed up my EA?


I have a QUAD CORE with 4 GIGs of memory. I am sure it is not the pc.


Every time my EA learns, it takes more memory space and disk space.

And eventually, it will learn everything but I know don't have the memory or the disk space.


Any good solution???????

 

code pls.

we need to see the code and give comments.

 
The only thing I can think of is compress combine and binary tree search what learned. Also throw out what is learned and then not used.
 

Thanks for information.

Binary Tree ain't bad.

The way I did was thru a fast hash index base upon key values.

As time went on, the knowledge base got larger and larger and larger.

I separated useful sets of data away from my main database linked by key fields into different files.

This improved the speed a little because it was no longer all stacked up inside one central database.

Even then, useful data sets still get larger and larger.

I mean the results are astounding! All currencies profits are climbing through the roof!

However, for every year of data learn means every hour of delays.

My EA learned so much that it just died. It froze and was not able to do anything anymore.

I believe it ran out of memory space and or disk space too.

If I can solve this dilemma, I want to add my AI to my current EA project.

And man, I feel real good about my current EA.

If my EA was combined with my AI, I know for a fact that mountains will move by themselve!

That is how good my EA and my AI will be when combined.

The issues is about NOT repeating the slowness and freezing problem.

 
The SLOWNESS may actually be MT4. It probably wants to cache the EA and other stuff into memory or something.
 

Memory management can be a real Sod with Microsnots progamming and the Intel chip doesn't help. Pity the transputer was too expensive for true parallel programming to get its foot in the door. I really wish they would use servers based on the language OCCAM programming language then we would have a really fast Internet. Wishful thinking - so far.

Your AI wouldn't have any problems in a computer built with a transputer.

 

CK

The space of a disk but the speed of RAM

http://www.lmgtfy.com/?q=ram+drive

Good Luck

-BB-

 

Hi CK/All,

I am a newbie in EA programming but have some experience in VB.

If you don't mind sharing your code... and your problem.

there might be a simple coding technic to solve this issue.

thanks.

 

Hi

try with 64bits windows and 16Go or more memory !

 
I'd suggest running an independent program for the AI and just have the EA communicate current prices and handle open and closes. The AI can run continuously, multithreaded and handle its own memory management.
 

You could try working with a lower level language. If you are using C++ it is possible to view the lower level C code
and then spend hours trying to improve on its effiency.
Or decide to learn in smaller chuncks and look to abstract specific learning into a higher more abstract category
Or look for a way of combining your learning method with a neural net.
Make sure that you are releasing objects created and have not used and check for memory leak.
That is object created and then released does not always give you back all the memory it used creating
memory leak. This is one of the reasons for coputers slowing down and needing a restart.
Im out of ideas now.