Machine learning in trading: theory, models, practice and algo-trading - page 418

 
Mihail Marchukajtes:

what does forex have to do with anything???? about forex read my previous posts or the article..... I'm not talking about forex to you....


So I'm off to read the classics.

 
elibrarius:
If you retrain, there will be no error (this is not retraining, because the coefficients will be reset) and just find their new combination.


Have you tried ensembles? It's true that bagging isn't used there, and they say here that boosting rules...

Again, I don't understand how to set up the ensemble in the Alglib, I have no help here, as usual... ) I'll have to look into the code.

http://alglib.sources.ru/dataanalysis/mlpensembles.php The main advantage, as written, I don't need to bother with network architecture, it's enough to create an excessive number of neurons, and the network itself will find optimal number of neurons.

 
Maxim Dmitrievsky:


Have you tried ensembles? They don't use bunching but bagging, but here they say that bunching works...

Again, I didn't understand how to set up the ensemble in the Alglib, help is nil as usual. ) I'll have to figure it out in code.

http://alglib.sources.ru/dataanalysis/mlpensembles.php The main advantage is that you don't need to bother with network architecture, you just need to create an excessive number of neurons and then the network itself will choose the optimal number

I tried it - it works. It takes longer. But I don't know which is better. I haven't made mass tests with comparisons yet... Besides, it takes a lot of machine time, and my own time too...
 
it all comes down to one thing. man wants to be god, hence the desire to create and teach something (machines, for example), and since we are machines ourselves, is it not easier to study ourselves and build something in our own image and likeness, as someone did it before us, having incarnated us in this world? Why invent something else, if this "else" already exists and it needs only to be realized. and it can be realized already now and everyone does it with pleasure, creating the pseudo-copy in the children, put in the child the algorithm necessary for you by training it to necessary things and you will receive the OWN Miracle. I don't understand your desire to put some algorithms into a car when you have a perfect, ready-to-use, understandable to everyone, tool to produce your copy.Teaching machines you refuse to teach your children, giving them much more time than your children, hence we get clumsy kids who run around the yards and break windows, deprived of fatherly attention and seeking that same attention on the side.
c/o.
 
Andrey Kisselyov:
in its own image,

So neural networks are built in the image and likeness of the human brain. So it's okay...

 
elibrarius:
Tried it - it works. Longer. But what is better - I do not know. Mass tests with comparisons have not done yet... Besides, it takes a lot of machine time, and my own too...

And how to do it, could you please send me an example? I understand, you need to create multiple copies of mLp and send them to the ensemble as parameters?
 
Andrey Kisselyov:
everything comes down to one. man wants to be god, hence the desire to create and teach something (for example, machines), and since we are machines, it is not easier to study ourselves and build something in our image and likeness, as who did it before us, embodying us in this world? Why invent something else, if this "else" already exists and it needs only to be realized. and it can be realized already now and everyone does it with pleasure, creating the pseudo-copy in the children, put in the child the algorithm necessary for you by training it to necessary things and you will receive the OWN Miracle. I don't understand your desire to put some algorithms into a car when you have a perfect, ready-to-use, understandable to everyone, tool to produce your copy.Teaching machines you refuse to teach your children, giving them much more time than your children, hence we get clumsy kids who run around the yards and break windows, deprived of fatherly attention and seeking that same attention on the side.
c/o.

+100500, I approve of what was said, everything is true, it really is.


Thanks for sharing :)

 
Maxim Dmitrievsky:

And how to do it, could you please send me an example? I understand, you need to create several copies of MLP and pass them to the ensemble as parameters?

You create a network, and then

CAlglib::MLPECreateFromNetwork( Net, ensemblesize, Ens);

 
elibrarius:

You create a network, and then

CAlglib::MLPECreateFromNetwork( Net, ensemblesize, Ens);

Ah, there you go, mani senx )
 

This mad troll now in person continues to write to me and now rubs his paid training courses and books on psychology :) type is clearly not all at home

So if anyone else begins to write - ban in advance, such rottenness ppts.