Feature custom llh

[yannikschaelte]

Allow to specify different cost functions (not only a normal distribution, the former only thing), normal staying the default.

So far, 6 fixed distributions (normal, laplace, on lin/log/log10 scale) are supported.

[codecov]

Codecov Report

Merging #656 into develop will increase coverage by 0.04%.
The diff coverage is 90.62%.

@@             Coverage Diff             @@
##           develop     #656      +/-   ##
===========================================
+ Coverage     72.9%   72.94%   +0.04%     
===========================================
  Files           48       48              
  Lines         6946     6972      +26     
===========================================
+ Hits          5064     5086      +22     
- Misses        1882     1886       +4

Flag	Coverage Δ
#cpp	69.49% <ø> (ø)	⬆️
#python	83.49% <90.62%> (+0.01%)	⬆️

Impacted Files	Coverage Δ
python/amici/sbml_import.py	82.27% <90.62%> (+0.43%)	⬆️
python/amici/__init__.py	83.63% <0%> (-1.82%)	⬇️

---

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 5bd7712...034390b. Read the comment docs.

[dweindl]

Suggested change

	# (no problem here, bu usually an upstream bug)
	# (no problem here, but usually an upstream bug)

[FFroehlich]

Would be awesome to have tests, not sure wheter amici will be able to compute dJydy for all of those functions,.

[yannikschaelte]

@FFroehlich Can you have another look? I added a basic test now (can be improved if that is better). Indeed, testing proves useful, because the abs terms beforehand did not work. Now it works (tested also on a few own models).
I had to tell sympy that the needed variables are all real-valued to get the correct derivative of abs -> sign.
But I think that makes sense anyway?

[FFroehlich]

Awesome, looks good!