Feature/dim reduction

[ots22]

Adds a (python-only) implementation of the gKDR algorithm for dimension reduction

[edaub]

A few comments, mostly it looks good.

I would explicitly check the value of K, as it is potentially ambiguous where the problem is if it has a bad value as it won't raise an error until you call __call__ even though the problem arose on calling __init__. I think a negative value should still run to completion without incident; however it leads to a situation where you specify the number of dimensions to drop rather than keep. If you want to allow that I might make that behavior more explicit in the documentation.

Relatedly, you may also want to provide exposed get/set interfaces to K, as it can be changed independently of the solving routine to generate the reduced dimension space. Users may not know the number of dimensions at the start, so this might be a nice way to let them easily play around with the choice without having it perform the full calculation each time.

You may also want to check positivity on EPS -- it might be unclear what the problem is if it has a large enough negative value to produce a singular matrix.

You divide by SGX2 (and SGY2 when forming the gram matrix), you may want to check these are not zero. Though the cause of this is likely to be clear enough so up to you how to handle this.

Minor thing: I have noticed that if my docstrings contain backslashes for LaTeX typesetting, Python will sometimes complain through a deprecation warning. This is fixed by making the docstrings raw strings.

[codecov-io]

Codecov Report

Merging #27 into devel will increase coverage by 0.36%.
The diff coverage is 100%.

@@            Coverage Diff            @@
##           devel      #27      +/-   ##
=========================================
+ Coverage   86.7%   87.07%   +0.36%     
=========================================
  Files         17       19       +2     
  Lines       3429     3527      +98     
=========================================
+ Hits        2973     3071      +98     
  Misses       456      456

Impacted Files	Coverage Δ
mogp_emulator/DimensionReduction.py	100% <100%> (ø)
mogp_emulator/__init__.py	100% <100%> (ø)	⬆️
mogp_emulator/tests/test_DimensionReduction.py	100% <100%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 0cdd4aa...ea7b974. Read the comment docs.

[ots22]

Thanks for the comments @edaub - added some checks. I'll leave setting K out of the interface for now.