Machine learning in trading: theory, models, practice and algo-trading - page 976
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
I wonder why this may happen?
I use DNN Darch, backpropagation, dropout.
Epoch: 16 of 100
Classification error on Train set: 41.69% (2668/6400)
Epoch: 17 of 100
Classification error on Train set: 31.87% (2040/6400)
Epoch: 18 of 100
Classification error on Train set: 32.09% (2054/6400)
Epoch: 19 of 100
Classification error on Train set: 39.55% (2531/6400)
Epoch: 20 of 100
Classification error on Train set: 38.02% (2433/6400)
Epoch: 21 of 100
Classification error on Train set: 49.89% (3193/6400)
Epoch: 22 of 100
Classification error on Train set: 50.56% (3236/6400)
Epoch: 23 of 100
Classification error on Train set: 45.56% (2916/6400)
That is, it takes a well-trained network from step 17 and degrades it by step 21.
Got it.
Didn't help:I put a new mask on each epoch, I'll switch to each miniBatch - should get better
Classification error on Train set: 45.11% (2887/6400)
Epoch: 10 of 100
Classification error on Train set: 34.92% (2235/6400)
Epoch: 11 of 100
Classification error on Train set: 30.16% (1930/6400)
Epoch: 12 of 100
Classification error on Train set: 45.28% (2898/6400)
Epoch: 13 of 100
Classification error on Train set: 39.56% (2532/6400)
That's fine. How should it be? After training you will get back the best model. You choose the criterion of "betterness" by yourself. You know that.
Reducing the batch does not always result in better quality.
Is it a model with optimized hyperparameters? Show me your model parameters and your training scheme. Maybe something will become clear.
Good luck
This is normal. How should it be? After training, the best model will be returned to you. The criterion for "better" is up to you. Well, you know that.
Reducing the batch does not always result in better quality.
Is it a model with optimized hyperparameters? Show me your model parameters and your training scheme. Maybe something will become clear.
Good luck
Just on your charts error jumped by 5 percent after approaching 30%. I got 30, I got 50.
Yes, the best model will get out.
And yes - such models are obtained when optimizing hyperparameters.
I tend to think that it's LearnRate = 1. But I tried 0.7 and 0.3 and 0.1 - and 0.01. Somewhere up to 0.1 is still something with an error of 30%, and at low LearnRate <0.1 NS can not learn anything at all. So at LearnRate = 1 let it find several variants from which I will then take the best one.
Second day struggling with Anaconda. In two days, already three reinstallations. When initializing Spyder, I get errors.
Yesterday everything was OK. Installed numpy, the cat should already be in the base, but it was not. That's when errors started.
I uninstalled numpy and installed it again with no effect.
Next write import numpy - the answer is that the module is installed but not initialized.
Reinstalling Anaconda with cleaning registry and removing all remnants of Anaconda related files and folders didn't help.
I'm freaking out.))
Second day struggling with Anaconda. In two days, already three reinstallations. When initializing Spyder, I get errors.
Yesterday everything was OK. Installed numpy, the cat should already be in the base, but it was not. That's when errors started.
I uninstalled numpy and installed it again with no effect.
Next write import numpy - the answer is that the module is installed but not initialized.
Reinstalling Anaconda with cleaning registry and deleting all remnants of Anaconda related files and folders didn't help.
I'm amazed.))
why do you need conda, first learn the basics of python, then IPython, then conda which is not necessary at all
The same Tflow and sklearn are installed without it
I'm amazed.))
Learn the basics of computers. With the computer, so to speak.
Learn the basics of computers. With a computer, so to speak...
))
Learn the basics of computers. A computer, so to speak...
Can you do it yourself? I doubt it).
IPython Desktop Reference Guide
https://jakevdp.github.io/PythonDataScienceHandbook/
Why do you need conda, first learn the basics of python, then IPython, then conda, which is not necessary at all
The same Tflow and sklearn can easily be installed without it.
I am aware of that. I liked Spyder. The standard Python one has little functionality. And Spyder is almost VS (just kidding). I must say I haven't seen any other editors besides the standard one yet.
As for the rest, everything is fine with Python - it works without problems, but there is some confusion with Spyder. And it already glitches out of the box, and only with numpy and matplotlib. The rest is still working, but I haven't gone far with Spyder-ra yet.
I don't want to install/try all editors in a row, either.