Parallel Processing

[shane-kercheval]

I imagine there are significant opportunities to speed up code via parallelization.

http://sebastianraschka.com/Articles/2014_multiprocessing.html http://www.paulbrownmagic.com/blog/python_functional_iteration/

• [x] ModelAggregator
• [x] ModelTuner
• [x] ModelSearcher
• [x] Resampler (this is quite difficult because there are various spots (e.g. callback) that don't fit into the parallelization scheme e.g. don't pickle)
• [ ] ModelStacker _(this is more complicated than it appears (or compared with Aggregator) because of the "blending" of model predictions into the final train_meta data-frame)_
• [x] update Jupyter notebook examples

Note: some of the above may be solved through others (e.g. ModelSearcher via Tuner)

[shane-kercheval]

it seems my implementation is not true parallel processing, despite the apparent gain i was seeing (false positive)

https://stackoverflow.com/questions/26432411/multiprocessing-dummy-in-python-is-not-utilising-100-cpu

[shane-kercheval]

Not implementing for ModelStacker, perhaps at a later date