A quick and free library for MT4, much to the delight of neuralnetworkers - page 43

 

I opened a trade with my hands, we'll see how it opens automatically after it closes, but I highly doubt it will. There are two possibilities, either the DC is messing up or the vpc is being stupid.

 

maybe a problem with the submitted transaction volume?

what is the real account? does it allow a trade of 0.02 lots?

 
umm. in the settings of the EA, the General tab, enable it to open trades in both directions and uncheck the option to confirm trades manually. and other stuff like that. as i remember, such an EA should work on small timeframes. and see what values the ST gives after we give it the pattern for processing. with proper training, this mismatch should be within 4-5% of the value you fed the ST during training
 
progma137:

maybe a problem with the submitted transaction volume?

what is the real account? does it allow for a 0.02 lot volume?

Yeah, it's from 0.01, I opened a small account specifically for this EA. I also used Alpari. I opened it with 0.02.

 
progma137:
Do you want to see what happens when you feed the robot to the NS?


I checked all the settings, I know MT5 for about five years.) My Expert Advisor is working on n1, like the author's, there is a branch here.

"Look at the values given out by the NS after we send it the pattern for processing. with proper training, this mismatch should be within 4-5% of the value that you fed the NS during training" - I do not understand this point ....

 
marker:


I've been familiar with mt5 for five years.

I'm confused by this phrase.


For example, I've done the experience. and I can recommend it to you.

so, 1 ns, not 15 as in your case. NS type (4,3,8,5,1)

neurons (four-layer network): 3 at the input, 8 first internal, 5 second internal and 1 at the output

there is data for the input - a cubic matrix consisting of zeros and ones. the matrix has 1000 cells

I draw some nonsense in the matrix. for example, a square of ones and fill everything around it with zeros.

I randomly pass all the cell values to train the NS.

For example, 4,0,2,1 says that cell number 4 (by count 5, x-axis), then cell number 0 (by count 1, y-axis) and the cell number 2 (by count 3ya, zet axis) is one.

after training, I run the train function through NS several times, until MSE is less than my user one, or when the result of using NS reaches 100% of correct results (it happened)

when we feed our pattern (coordinates, so the NS can guess what's inside) to the NS, it returns a single number of type double

in my case, for cells with zeros, the interval was -0.04 to 0.05 (the nitro swirled around zero but rarely got to zero)

cells with zeros ranging from 0.95 to 1, but not more than 1

Later i noticed, that cells filled with zeros, NS defined as 1, only because a particular cell was close to a cell with the value 1. but such interesting situations were very few. in other words, NS tried to magnetize or something, or show that we are close to the truth

 

I don't get it, I'm not a proger, I just use mt4:) Not mt5, mt4 of course....

 

So I don't understand how to make it work on the wpc.

 

How do I get it to work on seven?

 
marker:

How do I get it to work on seven?

Lack of a specific question, and the resulting lack of answers.... )

With this broker on this version of the terminal on WinXP - does it work?