Machine learning in trading: theory, models, practice and algo-trading - page 1283

 
elibrarius:
I will not make a release from one line. I understand that other people's time is less valuable than my own. And I take exactly the same position.
I showed a simple solution. Who needs it - he will repeat it.

And here I do not quite understand, if you did and showed the main part of the code edits, why not just lay out what is, or there are other edits, which simply do not have the time and desire to remove?

 
Aleksey Vyazmikin:

And here I do not quite understand, if you have done and showed the main part of the code edits, then why not just lay out what is, or there are other edits that just do not have time and desire to remove?

No time - that's right. Even reading the forum is distracting from more useful things. And you about the preparation of the release ... Well, it's a few hours to properly design and provide and explain, and then ask for support))

You are a programmer - as a programmer I am much more comfortable when it is clear what and why. Get into the code. In a couple of hours you'll understand the function of building a tree and shove the limit on the number of examples into it, and whatever you want.

Gone is the overkill of predictors to do.

 
elibrarius:

No time - that's right. Even reading the forum distracts from more useful things. And you about the preparation of the release... It's a few hours to properly design and provide and explain, and then ask for support))

You're a programmer - for me as a programmer it's much more convenient when it's clear what and why. Get into the code. You'll understand the tree-building f-fi and you can plug in both limitation on number of examples and whatever you want.

Gone to do the overkill on predictors.

By the way, off-topic, but... without serialization the binary loads much faster (instantly). I switched to serialization for some reason and agonized, switched back to direct saving. I could not teach it to read binary using serialization. So I understood, that only serial tostring works in alglib and I need intermediate conversion.

 
Maxim Dmitrievsky:

By the way, off-topic, but... without serialization the binary loads much faster. For some reason, I switched to serialization and had a hard time, so I switched back to direct saving

I liked it: no serialization. (Very much in tune.))

 
Yuriy Asaulenko:

I liked it - no serialization

What? There is "serialization" in the string and not in the bin

 
Maxim Dmitrievsky:

with what?

Haven't you ever heard of demons?

 
Yuriy Asaulenko:

Haven't you ever heard of demons?

I don't know about it or am I dumb?

ah... all ) used to read without looking, the brain compresses the information, yo no... do not pay attention to errors
 

https://www.kaggle.com/itoeiji/deep-reinforcement-learning-on-stock-data

interesting study on RL in python. Except that the train and test pieces are almost the same ) so in profit

I have a similar ideology now ... look for "good" chunks and trade only on them
 
Maxim Dmitrievsky:

So I understand that in the alglibe only the series works

Yes, and to make your own save/read forest into binary - no problem ( FileWriteArray or FileWriteStruct). I have not done it yet, but I think it will take no more than 2 hours (together with testing).
 
elibrarius:
Yes, and it's no problem to make your own save/read forest into binary ( FileWriteArray or FileWriteStruct). I haven't done it yet, but I think it won't take more than 2 hours (together with testing).

my 1st article has "fast" saving and loading via binary. the 2nd article has slow loading via deserialization, the saving speed does not seem to suffer