Machine learning in trading: theory, models, practice and algo-trading - page 459

 

Another thing is that the power does not allow to train files of larger size, it takes too long.... That's why I'm saving up for a good computer, 20-30 cores, and there we'll see....

 
Mihail Marchukajtes:

Another thing is that the power does not allow to train files of larger size, it takes too long.... That's why I'm saving for a good computer, 20-30 cores, and then we'll see....


Yes, very slow, so I gave up on it, and not convenient for expanding the system and the selection of predictors in the optimizer

so it's cool but hard to use

 
Maxim Dmitrievsky:


yes, very slow, that's why i gave up on it, and it's not convenient for expanding the system and picking up predictors in the optimizer

so it's cool but hard to use


I agree, but the data is shaken up up all over the place, and more than once... so... All means are good in war, the main thing is the final result....

 
Mihail Marchukajtes:

I agree, but the data is shaken up up all over the place, and more than once... so... All means are good in war, the main thing is the end result....

One more nuance - try it on crypto, on bitcoins) a very predictable stable trend, on forex is worse
 
Maxim Dmitrievsky:
One more nuance - try it with crypto, with bitcoins they have very predictable stable trends.

I read Reshetov's posts about volatility and it turns out that it is quite simple to predict the volatility and it is almost risk-free to trade with it. Market-neutral strategies.... Very intrigued, I think to look at it at my leisure ....

I think that at crypto I understand that trends do not really exist, changes happen in leaps and bounds. I THINK THERE IS NO TREND ON CRYPTO.

 
Mihail Marchukajtes:

I read Reshetov's posts about volatility and it turns out that it is quite simple to predict the volatility and it is almost risk-free to trade with it. Market-neutral strategies.... Very intrigued, I think to look at it at my leisure ....

I think that at crypto I understand that trends do not really exist, changes happen in leaps and bounds. IMHO.

No, all is normal there... or some indexes or shares that constantly go up... starting to fall - disconnect neuronics.)
 
Vizard_:

Tell me how to build a normal model for the Eurodollar. With examples.

You don't give a shit and you're all hilarious. All you do is avoid arguing your claims. This is another evasion. Until you tell us about the two mistakes in Reshetov's RNN, there's nothing to talk about. For that is trolling.
 
Mihail Marchukajtes:

I read Reshetov's posts about volatility and it turns out that it is quite simple to predict the volatility and it is almost risk-free to trade with it. Market-neutral strategies.... Very intrigued, I think to look at it at my leisure ....

I think that at crypto I understand that trends do not really exist, changes happen in leaps and bounds. IMHO.


It's called GARCH. It's an extremely developed thing, widely used in financial markets, including Forex. Consider it mainstream.

 
Vizard_:

It sucks when you can't understand what you're talking about, even after Doc wrote and Misha confirmed...


You're beating around the bush...... You tell me specifically, give me an example... Then we can continue the conversation, but otherwise....

 

And who tried to take the simplest thing: a single-layer percetron input above zero, output below, and then add another one and input again if the first and second are above zero, etc.?

In this case the output will be carried out, only on the basis of the first percetron.

The neural network should be used if you understand how it is set up. So, how to modify, to try different variations, and how to do it, if you do not understand the principle of its operation?

Reason: