Machine learning in trading: theory, models, practice and algo-trading - page 2684

New comment
 
Maxim Dmitrievsky #:
There is a TCN for time series
Where.
 
mytarmailS #:
Where?
I don't know where, it's an algorithm, or TCNN, something like that.
Time series convolutional NN
 
Maxim Dmitrievsky #:
I don't know where, it's an algorithm, or TCNN, something like that.
Time series convolutional NN
Mm-hmm, got it. It's used for speech
 
mytarmailS #:
Yeah, got it. It's for speech.

Well, transformers are also for letters, mostly, depending on the traits.

 
Maxim Dmitrievsky #:

Well, Transformers are also for letters, mostly, depending on the traits.

Yeah, sequences...

Shit... Looking at examples on torch, to write a neuron you have to write dozens of lines of code, it's unusual after R with two lines.
 
mytarmailS #:
Yeah, sequences...

Shit... Looking at examples on torch, to write a neuron you need to write dozens of lines of code there, it's unusual after R with two lines.
PyTorch is a cool constructor, you just need to understand it. Then you can build any kind of architecture

But do you need it? Take the unrivalled catbust and that's it 😀

I compared neural networks with bousting for my tasks. Same output, but bousting is faster to learn.
 
Maxim Dmitrievsky #:
PyTorch is a cool constructor, you just need to understand it. Then you can build any architecture you want

But what do you need? Take the unrivalled catbust and that's it 😀

I compared neural networks with bousting for my tasks. Same output, but bousting is faster to learn.
On MNIST At least try comparing bousting and CNNs
 
mytarmailS #:
On MNIST, at least try to compare the boost to CNN.
There's nothing to do. The difference is usually +-. On some specific datasets like signal processing neural networks are brighter. But we found out yesterday that there are no signals on Forex 😀
 
Maxim Dmitrievsky #:
You're just jumping from one thing to the next.
You're right about that. Why go deeper? If there's not even a hint of slavery? I tried it, it's no good, it's the next one.
 
mytarmailS #:
You're right, of course. Why bother going into it? If there's not even a hint of workability. I tried it, it's no good.
That's normal in forex. People are always tinkering with algorithms, sometimes they get lucky.
1...267726782679268026812682268326842685268626872688268926902691...3547
New comment
Reason: