Adding simple baselines from GIS

[patel-zeel]

Simple deterministic baselines from geographical information systems (GIS) such as "Inverse Distance Weighting", "K Nearest Neighbors", "Mean of the context y" act as good baselines and sometimes provide insights into the complexity of the dataset. It may be helpful to implement these baselines inheriting from DeepSensorModel class.

[tom-andersson]

Adding baselines sounds great @patel-zeel. Do you envisage having a new module deepsensor.model.baselines implementing each of these?

[patel-zeel]

@tom-andersson Grouping similar characteristic models seems a good idea. Can we use a more self-explanatory name for this module? For example,

• deepsensor.model.gis or deepsensor.model.deterministic or deepsensor.model.interpolators
  • KNN
  • IDW
  • Kriging
  • SpatialAverage

Over time, we may target most of the interpolators from Esri's ArcGIS. Should we put a dependency on polire (a spatial interpolation library from our lab) for this to avoid reinventing the wheel?