GEKPLS With Custom PLS Based on SKLearn PLS

[vikram-s-narayan]

This version of GEKPLS replaces the SKLearn PLS in the previous version with a modified PLS version based on the SKLearn PLS code. I have also added more tests, additional functionality and documentation.

It does not include code equivalent to SMT's hyperparameter optimization for theta. I plan to work on hyperparameter optimization next as I'm still researching how best to implement it. I intend to submit a separate pull request for hyperparameter optimization.

[codecov]

Codecov Report

Merging #359 (cd72b3f) into master (f7d2447) will increase coverage by 1.30%.
The diff coverage is 92.16%.

@@            Coverage Diff             @@
##           master     #359      +/-   ##
==========================================
+ Coverage   77.59%   78.89%   +1.30%     
==========================================
  Files          15       16       +1     
  Lines        2053     2270     +217     
==========================================
+ Hits         1593     1791     +198     
- Misses        460      479      +19     

Impacted Files	Coverage Δ
src/Surrogates.jl	18.18% <ø> (ø)
src/GEKPLS.jl	92.16% <92.16%> (ø)
src/Optimization.jl	72.16% <0.00%> (-0.19%)	⬇️

📣 Codecov can now indicate which changes are the most critical in Pull Requests. Learn more

[vikram-s-narayan]

Also, I ran a quick comparison between GEKPLS, Kriging and Radial Basis.

Results for the water flow function are below:

rmse_rad: 50.399
rmse_krig: 46.086
rmse_gekpls: 0.0185

[ChrisRackauckas]

Wow that's awesome!

[ranjanan]

nice! Also compare PolyChaos? like in the docs: http://surrogates.sciml.ai/stable/water_flow/? Curious how that stacks up as well.

[vikram-s-narayan]

nice! Also compare PolyChaos? like in the docs: http://surrogates.sciml.ai/stable/water_flow/? Curious how that stacks up as well.

rmse_poly: 1.518

Updated gist here.

[ChrisRackauckas]

Set some strict modes in the doc build: https://github.com/SciML/DiffEqSensitivity.jl/blob/v6.79.0/docs/make.jl#L16-L23

[ChrisRackauckas]

@ranjanan are you doing the review?

[ranjanan]

Ah review didn't get posted. Posted now.

[vikram-s-narayan]

Set some strict modes in the doc build: https://github.com/SciML/DiffEqSensitivity.jl/blob/v6.79.0/docs/make.jl#L16-L23

I'm currently getting an error ("ERROR: makedocs encountered an error (:docs_block, :example_block). Terminating build before rendering.") without being able to create the build directory when I make the following change in make.jl:

makedocs(
    sitename="Surrogates.jl",

    strict = [
             :doctest,
             :linkcheck,
             :parse_error,
             :example_block,
             # Other available options are
             # :autodocs_block, :cross_references, :docs_block, :eval_block, :example_block, :footnote, :meta_block, :missing_docs, :setup_block
         ],
         
    format = Documenter.HTML(analytics = "UA-90474609-3",
                         assets = ["assets/favicon.ico"],
                         canonical="https://surrogates.sciml.ai/stable/"),
    pages = pages
)

So my understanding is that each page in docs will need to be fixed independently in order to make these tests pass. I will work on these separately and send in other pull requests for documentation not related to GEKPLS.

[ChrisRackauckas]

So my understanding is that each page in docs will need to be fixed independently in order to make these tests pass. I will work on these separately and send in other pull requests for documentation not related to GEKPLS.

Yes that's fine.

Note you may need to run the formatter to fix the conflicts. Just do:

using JuliaFormatter, Surrogates
format(joinpath(dirname(pathof(Surrogates)), ".."))

[ChrisRackauckas]

I think this is generally good to merge after conflicts are handled. We can do the docs stuff in a separate PR. And we should open an issue to profile and optimize the implementation, but I don't think there's any major blocking in the code here.

[vikram-s-narayan]

I ran the JuliaFormatter but the format-check still fails. Assuming that I've run the formatting code correctly, my guess is that this format-check fail is happening because newly added and modified Julia files will always show up in the git diff

So the condition in the following code may need to be altered?

  julia -e '
  out = Cmd(`git diff --name-only`) |> read |> String
  if out == ""
      exit(0)
  else
      @error "Some files have not been formatted !!!"
      write(stdout, out)
      exit(1)
  end'

[ChrisRackauckas]

No. You ran the formatter before you rebased to master, so you ran it with the wrong style because it didn't have the JuliaFormatter.toml file in there. Revert that one (because it did a bunch of spacing changes that aren't preferred but aren't against the style, so they would be kept) and run the formatter on a post-rebased commit.

[ChrisRackauckas]

This shouldn't be required in the Project?

[ChrisRackauckas]

🎉 thanks for seeing this through, this is a huge step.

[ChrisRackauckas]

oh I noticed the the formatter CI is gone. I'll just add it back and handle any issue there.