Add comment

[willtebbutt]

Addresses #67

[codecov]

Codecov Report

Merging #68 (948912b) into master (d398b73) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master      #68   +/-   ##
=======================================
  Coverage   86.99%   86.99%           
=======================================
  Files           5        5           
  Lines         246      246           
=======================================
  Hits          214      214           
  Misses         32       32           

Impacted Files	Coverage Δ
src/svgp.jl	93.18% <ø> (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 d398b73...948912b. Read the comment docs.

[theogf]

BTW any reason SVGP is not an AbstractGP ?

[st--]

BTW any reason SVGP is not an AbstractGP ?

Mhmm, I see it as more like LaplaceApproximation() than GP(). A way of specifying an approximation, not a GP in itself (the GP itself you get by calling posterior(...) on it)

[willtebbutt]

Agreed @st-- . This might be a good candidate for re-naming though, because SVGP is literally "sparse variational Gaussian Process", which makes it sound an awful lot like it ought to be a GP. Perhaps we should go with something like SparseVariationalApproximation

[rossviljoen]

I'm in favour of renaming to SparseVariationalApproximation

[willtebbutt]

Excellent. I've opened in a new issue so that we don't forget. @theogf is there anything else, or are you happy for this to approve?