rlberry icon indicating copy to clipboard operation
rlberry copied to clipboard

Replace scikit-fda by handmade Nadaraya-Watson smoothing

Open TimotheeMathieu opened this issue 8 months ago • 13 comments

Description

In this PR I code by hand the Nadaraya-Watson smoothing. The disadvantage is that I am less sophisticated and I don't unse cross-validation to estimate the bandwidth. The advantage is that we do not depend on scikit-fda anymore.

Still todo:

  • [x] add more explanation about bandwidth parameter
  • [x] return the smoothed curves as pandas to be able to custom plot them

TimotheeMathieu avatar Mar 21 '25 16:03 TimotheeMathieu

Hello, The build of the doc succeeded. The documentation preview is available here: https://rlberry-py.github.io/rlberry/preview_pr

github-actions[bot] avatar Mar 21 '25 16:03 github-actions[bot]

Hello, The build of the doc succeeded. The documentation preview is available here: https://rlberry-py.github.io/rlberry/preview_pr

github-actions[bot] avatar Mar 23 '25 13:03 github-actions[bot]

Hello, The build of the doc succeeded. The documentation preview is available here: https://rlberry-py.github.io/rlberry/preview_pr

github-actions[bot] avatar Mar 24 '25 13:03 github-actions[bot]

Hello, The build of the doc succeeded. The documentation preview is available here: https://rlberry-py.github.io/rlberry/preview_pr

github-actions[bot] avatar Mar 24 '25 13:03 github-actions[bot]

Hello, The build of the doc succeeded. The documentation preview is available here: https://rlberry-py.github.io/rlberry/preview_pr

github-actions[bot] avatar Mar 24 '25 14:03 github-actions[bot]

  • Make clearer names, e.g. no "ed" for plot_smoothed_curves() maybe call it "smooth_and_plot()".
  • Also maybe add the argument to change n_simu column name.
  • Refactor code to separate smoothing and plotting ?
  • I don't think we need the plot_synchorinized curves.

KohlerHECTOR avatar Mar 26 '25 12:03 KohlerHECTOR

Hello, The build of the doc succeeded. The documentation preview is available here: https://rlberry-py.github.io/rlberry/preview_pr

github-actions[bot] avatar Mar 26 '25 15:03 github-actions[bot]

Hello, The build of the doc succeeded. The documentation preview is available here: https://rlberry-py.github.io/rlberry/preview_pr

github-actions[bot] avatar Apr 29 '25 09:04 github-actions[bot]

Hello, The build of the doc succeeded. The documentation preview is available here: https://rlberry-py.github.io/rlberry/preview_pr

github-actions[bot] avatar Apr 29 '25 12:04 github-actions[bot]

Hello, The build of the doc succeeded. The documentation preview is available here: https://rlberry-py.github.io/rlberry/preview_pr

github-actions[bot] avatar Apr 29 '25 12:04 github-actions[bot]

TODO:

  • [x] num of simulations may not be a sequence. This implies some bugs. (+ regression.)
  • [ ] change name to smooth_NW_and_plot.

TimotheeMathieu avatar Apr 29 '25 13:04 TimotheeMathieu

Hello, The build of the doc succeeded. The documentation preview is available here: https://rlberry-py.github.io/rlberry/preview_pr

github-actions[bot] avatar Apr 30 '25 13:04 github-actions[bot]

Hello, The build of the doc succeeded. The documentation preview is available here: https://rlberry-py.github.io/rlberry/preview_pr

github-actions[bot] avatar Apr 30 '25 13:04 github-actions[bot]