Machine learning in trading: theory, models, practice and algo-trading - page 1968
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
So that's kind of what this layer does.
The problem is not compressing it, the problem is that there will be billions of sequences, 99.9% of them will be garbage.
That's the problem, before you compress it, you have to get rid of the garbage.
When you trash it, you don't have to compress anything.)
Well, yes, but there is no memory of the agent's previous actions, this is different
I'm going to read the binder again, and then I'll dig through the code.
For me it's not memory, it just looks like memory, it's new data of NS result for quantization, besides it's already broken down by clusters. And it turns out that the system seems to remember, but in fact it gets the result data at the beginning of the BP section. At the same time it is not repeated passes, and taking into account the forward movement of the data. The question is how it receives and how it takes into account. There can be an extra neuron for accounting, or weighted data.
https://cran.r-project.org/web/packages/glmdisc/vignettes/glmdisc.html
package for discretization (quantization) with respect to the target
Thanks, took a look at it. Playing random again, as I understand it? Have you tried it yourself? I didn't see how to save the partitioning conditions later in the file, to apply when new data comes in.
I spent all day on Quickcom, guys, it's such a pain in the ass that you wouldn't wish it on your enemy. I'm telling the truth, the options module is really a mess :-(. And the main thing is that there are a lot of places where you can build the volatility smile. I don't see any analytical program or web services, but none of them has parameters to calculate the smile. I do not need to look at it if it is described by three values. And the main thing with no one these values are not calculated alas. Only smartX counts but it is under the Wine is not installed, or they have a curve distribution, have already downloaded all the options and miss. I remember adjusting the display smarthat like the MT and showed his appearance dude, so he just flipped at how cool it started to look. This is crazy comrades. Run away from Quick as fast as you can. Such bullshit....
Quicksilver is hard.
Thank you, I looked it up. Another game of random, as I understand it? Have you tried it yourself? I didn't see how to save the partitioning conditions into a file to apply when new data comes in.
in tensorflow-in keras-in theano -in mxnet-in pyTorch- etc ....
I didn't seethe button-"convert millions of lines of code, create and train a convolutional deep network in mql!!! because I'm a trader and I want to"
But the whole world is using it, it must be a game of random
:)
but there is a button to delete everything on...
))))
how's it going over there?
))))
how's it going?
went away... now more than an hour of testing from early 2019 to this day, I'll send you a screenshot soon
You know, at almost every bar the neural network is retrained.