Observation model

[t-kimber]

Description

This PR deals with the observation model implemented in kinoml (theory and implementation)

Todos

• check observation model in the api for all 4 measurement types (and measurement units)
  • pIC50
  • pKi
  • pKd
  • %displ.
• adapt kinoml workflow (with theory and image)

Question

• Where should the explanation (the theory behind the observation model) live?

Status

• Ready to go

[t-kimber]

TO BE CHECKED!

In:

• https://github.com/openkinome/kinoml/blob/observation-model/kinoml/workflows/KinoML_workflow.ipynb,
• https://github.com/openkinome/kinoml/blob/observation-model/examples/ChEMBL.ipynb,
  if executing the optimization loop cell gives the following error
  RuntimeError: Found dtype Double but expected Float,
  add y = y.float()right before y = y.reshape(prediction.shape).

[jaimergp]

Why do we need to hardcode this? Don't we have the argument in the function? Also, why are we changing units now?

[t-kimber]

I need to speak with John for this observation model. I thought that this measurement type was in the clear, but I realized at the end of last week that something is fishy and I need to loop in John asap.

[jaimergp]

RuntimeError: Found dtype Double but expected Float

This should be fixed by last commits.