Machine learning in trading: theory, models, practice and algo-trading - page 2053
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
And this is how the catbust was trained on the same data (in 5 seconds)
52: learn: 0.7964708 test: 0.7848837 best: 0.7860866 (27) total: 604ms remaining: 5.09s
Source dataset:
Trained model (the second half of the trade is the test sample):
Not always, of course, depending on the sampling (and it's random, i.e., it needs oversampling). Sometimes like this:
34: learn: 0.5985972 test: 0.5915832 best: 0.5927856 (9) total: 437ms remaining: 5.81s
Maxim, I have a question, what are your values on the axes on the diagrams, and did you do a graph for the net convergence?
Maxim, I have a question, what are the values on your charts on the axes?
The number of trades, by y profit in pips
I have only to save the model in Metac and check it in his tester
The number of trades, by y profit in pips
If you remember the last time we were talking about the paternoster, I put it for training on the same day, it is still learning.
Remember the last time we talked about the wrenches? on the paternoster, put it to learn the same day, still learning. it seems like a long time.
(Is the network written in metatrader? ) I have already commented on this subject
network is written in metatrader? ) I have already commented on this topic
i have already commented on it in metatrader) - but this is a plus)))) it does not retrain), make a graph of your network errors, i would like to see)
And this is how the catbust was trained on the same data (in 5 seconds)
52: learn: 0.7964708 test: 0.7848837 best: 0.7860866 (27) total: 604ms remaining: 5.09s
Source dataset:
Trained model (the second half of the trade is the test sample):
Not always, of course, depending on the sampling (and it's random, i.e., it needs oversampling). Sometimes like this:
34: learn: 0.5985972 test: 0.5915832 best: 0.5927856 (9) total: 437ms remaining: 5.81s
correct result 0.59
you can't just sample a time series, it's not fischer's irises))))
you're looking into the future ... sampling is strictly for the track, first divide, then sample
not vice versa as you did
On the mete, but there is a plus)))) it does not retrain)), make a graph of the errors of your network, I would like to see)
you cannot use algorithms that take so long to learn... you may go gray.
the correct result is 0.59
you can't just sample time series, it's not Fisher's irises)))
you're looking into the future ... sampling can be strictly trace, first divide, then sample
not the other way around like you did
what do you mean correct result? these are errors for different datasets
Not time series are sampled, but labels. see the videosYou can't use algorithms that take so long to learn... you can go gray.
Akurashi I guess that's prediction accuracy? And logloss? There shouldn't be learning on test, and error should be the same regardless of number of passes? or -+ at least, but it shouldn't decrease
What do you mean the correct result? these are errors for different datasets
it's not the time series, but the labels. see the videosI understand correctly that you're training the network to predict time series, right?