Machine learning in trading: theory, models, practice and algo-trading - page 359
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
Write down the words.)) I want, if not a Bentley, then at least a Peugeot with an automatic).
Logan bucket :)
Then the question of retraining in all its glory
Re-training has become a bugbear in the topic.
There is nothing new, scary, or even unusual about overtraining. It is inherent in absolutely all models, including deterministic ones. Even Y=Ax+B can be retrained.
Typical consequence of wrong choice of inputs, the model itself and required accuracy not corresponding to the process. Say, an attempt to refine a nonlinear process with a linear model will cause the model to crumble. A crude model will give good but insufficient results - it seemed, "well, pussycat, a little more" (c).
I don't even know if I need to do any better, 20,000% in 2.5 months at opening prices on 5 minutes if I'm lucky... you throw in $1k and pre-order a Bentley. If you're not lucky, no big loss.)
Did they do that on Reshetov's neuron (RNN) again?
Yes, but there's not much left of it anymore.)
Soon I will compare it to MLP z alglib, not quite the right comparison will be true
And then there's nothing left but to switch completely to R
I don't even know if I need to do any better, 20,000% in 2.5 months at opening prices on 5 minutes if I'm lucky... you throw in $1k and pre-order a Bentley. If you're not lucky, no big loss.)
Maxim, here is a contest scheduled in July for 4 weeks - the question price of $ 10. Not a bad way to test in real life in what ever the real battle mode.
Maxim, here is a contest scheduled in July for 4 weeks - the question price of $ 10. Not a bad option on the real world to test in the most or at all a combat mode.
Maybe if there will be a final version by this time, so far I have a new version every day
Yeah, but there's not much left of it.)
Soon I'll be comparing it to MLP z algleeb, not really a fair comparison really.
And then there will be nothing left but to switch completely to R
Then the question of retraining in all its glory
Nowadays, many methods have been developed in deep neural networks to significantly reduce the probability of overfitting.
In general, the translation "overfitting -> overtraining" is inaccurate in meaning. It is more likely to mean "to learn" as opposed to "to learn".
You don't have to be afraid of it, you just have to control the moment when you finish learning.
Good luck
Maxim, here is a contest scheduled in July for 4 weeks - the question price is $10. It is a good way to test in real trading mode.
Maybe if there will be a final version by this time, so far I have a new version every day.
It's time! It's time already!