[MRG] accelerate plot_successive_halving_iterations.py example #21598#21612
Conversation
|
@sply88 could you please paste both before and after outputs of the example Your change is trigerring an issue in the doc build. The CI fails. |
Seems like |
|
@NicolasHug @rth could maybe help here? |
thomasjpfan
left a comment
There was a problem hiding this comment.
Syncing with upstream should fix the issue. I think it is related to #21607 and how numpy is being updated in the min-doc build when installing the latest version of PyWavelets
…ing_iterations.py
Works. Thanks for pointing that out @thomasjpfan |
| X, y = datasets.make_classification(n_samples=400, n_features=12, random_state=rng) | ||
|
|
||
| clf = RandomForestClassifier(n_estimators=20, random_state=rng) | ||
| clf = RandomForestClassifier(n_estimators=15, random_state=rng) |
There was a problem hiding this comment.
Can we leave this at 20? I like how the last iteration shows a difference between the last two candidates:
There was a problem hiding this comment.
I agree, result looks better with a unique winner. Only takes around 1s more on my machine. Thanks again for fine tuning!
…earn#21598 (scikit-learn#21612) * accelerate plot_successive_halving_iterations.py example scikit-learn#21598 * n_estimators back to 20
…earn#21598 (scikit-learn#21612) * accelerate plot_successive_halving_iterations.py example scikit-learn#21598 * n_estimators back to 20
…earn#21598 (scikit-learn#21612) * accelerate plot_successive_halving_iterations.py example scikit-learn#21598 * n_estimators back to 20
Speeds up ../examples/model_selection/plot_successive_halving_iterations.py (Issue #21598) by
n_estimatorsmax_featuresFor me example runs in 5 sec now (previously plus 13).
Reducing number of samples also reduces number of iterations during search (now 5, previously 6). Final figure:
Original figure:
