Machine learning in trading: theory, models, practice and algo-trading - page 202
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
Keep using R from mt5 via crutches. thanks to them - mt4r.
I'm sorry. Does mt4r for 5 also work? Have you checked it?
look here, youtz https://www.mql5.com/ru/forum/37931/page10#comment_2945898
Article "Computing discrete mixtures of continuous distributions" is available on the author's website.
The authors of the article say that the problem is in the convergence criterion of the series.
The implementation of their proposed recurrence algorithm 7.2 https://github.com/neurodebian/afni_removeme_eventually/blob/master/nct.c.
However, the recurrence calculation has errors. For example, for the beta function:
#include <Math\Stat\Math.mqh>
//+------------------------------------------------------------------+
//| Script program start function |
//+------------------------------------------------------------------+
void OnStart()
{
//---
double a=1;
double b=1;
double r_beta=MathBeta(a,b);
for(int j=0; j<40; j++)
{
if(j>0)
{
r_beta*=((a+j-1)/(a+b+j-1));
}
double beta=MathBeta(a+j,b);
PrintFormat("%d error=%5.20e",j,beta-r_beta);
}
}
CDF calculation may be accurate enough, but it may not be accurate enough for converting quantiles with great accuracy.
Therefore, we used direct summation without recurrence calculations in the algorithm for calculating quantiles.
Gentlemen, what is the argument for 100,500 pages? Who knows more about math? What difference does it make to you as a matter of principle if the function is defined in 0 or not, if the article says there is an error in R or not. It's as if they are discussing your child here and saying that he is full of shit, while you stand up for him with your whole chest.
Take it as a given and use built-in features in MT5, ask developers for new features, if something is missing, write your own. If you miss something, write your own. At the end of the day go on using R from mt5 through crutches, fortunately they exist - mt4r.
There is no fundamental difference. There is no difference in principle, and even less of a practical difference.
And there is - an answer to categorical clichés. And if there are no such answers, you can screw any standard (including the IEEE - the largest scientific community) and write whatever you want.
There is no difference in principle. There is no difference in principle, much less in practice.
And there is - an answer to categorical clichés. If there will be no such answers, then you can lay on any standard (including the IEEE - the largest scientific community) and write whatever you want.
You could not oppose anything against the explanation of @Quantum, which is very deeply researched the topic and showed the root of the error. He wrote the library, did in-depth research, showed the rationale, while all you did was ask questions and believe in R.
Our conclusions were correct.
Gentlemen, what is the argument for 100,500 pages? Who understands mathematics better? What difference does it make to you as a matter of principle- is the function defined in 0 - is not defined, has written in the article - error in R - or not error. as if here they discuss your child and say that he is complete ..., and you whole chest stood up to defend him.
There is no difference in principle. There's no difference in principle, much less in practicality.
I couldn't agree more. Maybe you won't encounter it in practical calculations, but the density must be defined correctly.
An example was found
which shows that you can't just return the limit value, otherwise the probability density is infinite. This is an error.
The correct way to define it is as follows:
Then everything is logical and correct.
In Matlab it is done the same way:
>> gamma_pdf(0.5,1,0)
ans =
0.00000000000000e+000
>> gamma_pdf(1,1,0)
ans =
0.00000000000000e+000
You couldn't counter anything against the explanations of @Quantum, who worked very deeply on the topic and showed the root of the error. He wrote the library, did in-depth research, showed the rationale, and on your part just questions and belief in R.
Our conclusions were correct.
The path you have chosen is clear. To build a toolkit which will allow (as you think) to solve the same problems as in R. It is possible.
But you must understand that MCL and R are languages of different abstraction level.
MCL is an implementation language with all that it implies. It forms the corresponding way of thinking for programmers.
R/Python are algorithmic languages, with a higher level of abstraction, allowing you to solve complex problems without having to worry about the level of implementation. Yes, sometimes bottlenecks appear when profiling complex calculations. So, let's go down, write a Cp file and move on. If that does not help, parallelize it, or connect the graphics board. And for all this, there are ready and tested solutions!
You propose to start everything from the bottom. I don't argue that your tools may be golden today, some of them you may even make platinum, but they will remain tools. Figuratively, having good tools and dizzy hands in the garage, you can build a bike, or you can take ready-made assemblies and build a bike with the desired properties. There is no point in arguing which of these approaches is better. They are different and both have the right to life. Everyone can choose according to their preferences.
But these are two roads in parallel planes, I doubt they will cross in the near future.
Good luck
I cannot agree here. You may not encounter this in practical calculations, but the density must be determined correctly.
An example was found
which shows that you can't just return a limit value, otherwise the probability density is infinite. This is a mistake.
The correct way to define it is this:
Then everything is logical and correct.
Matlab does the same thing:
OK, accepted.
And why do you say, "...correctly so..."? From what considerations is it correct? Poke me again at Wolfram, where statistician Uncle John wrote "otherwise 0". Your own considerations as an expert as to why it is correct to define density as 0 in zero? And examples of when defining density as 1 or inf becomes a problem? Without pointing to large software projects, please...
And I've already given examples of discussions where there were examples that 1 is more convenient... You can read my last post at the links.
Thanks.
The path you have chosen is clear. To build a toolkit which will allow (as you think) to solve the same problems as in R. Perhaps.
You suggest starting everything from the bottom. I don't argue that your tools today can be gold and you can even make some of them platinum, but they will remain tools. Figuratively, having good tools and dizzy hands in the garage, you can build a bike, or you can take ready-made assemblies and build a bike with the desired properties. There is no point in arguing which of these approaches is better. They are different and both have the right to life. Everyone can choose according to their preferences.But these are two roads in parallel planes and I doubt that they will intersect in the near future.
Our approach - port your developments from R to MQL5 using our standard libraries.
You will see that it is not only possible but easy. We have embarked on a major campaign to implement complex functions in libraries and the language itself.
Today we are going to publish a beta and show you another cool feature with graphical libraries to replace plot in R.