Plotting learning curve step takes REALLY long time

[dan-lind]

More of a question than an issue.

Is it to be expected that the "Calculating Learning Curve..." step takes a really long time run? I'm on a Macbook Pro with 4 cores, and the step took 20 minutes to complete.

[davenovelli]

I recall it taking quite a while for my quad-core lenovo as well, which I suspect is normal. You can reduce the number of cross validation folds to get a little bit of a speedup, but in any case it's training a lot of large random forests. You can also comment this step out if you're iterating a lot. The ROC curve calculation is a lot faster, and you could turn the learning curve back on as a check after finding good results with the ROC validation.

[mbaytas]

To monitor the progress of learning, you can fork the code and pass a verbose parameter to the learning_curve() function call in learningcurve.py.

See:

• http://scikit-learn.org/stable/modules/generated/sklearn.learning_curve.learning_curve.html
• http://stackoverflow.com/questions/29995249/verbose-argument-in-scikit-learn
• http://stats.stackexchange.com/questions/153823/what-is-verbose-in-scikit-learn-package-of-python
• http://stackoverflow.com/questions/22443041/predicting-how-long-an-scikit-learn-classification-will-take-to-run