hyperparameter tuning and ranges

[corybrunson]

Hyperparameter tuning requires that tunable parameters be identified for each step and assigned functions that generate suitable grid ranges. The current implementation provides hom_degree() and max_hom_degree(), which are internally the same and provide integer ranges between 0L and 3L.

Community input should be gathered (and literature reviewed) to describe best practices and design range generators accordingly. This issue should remain open until all potentially tunable parameters are decided upon in the checklist below. Each should be decided "No" or "Yes", with default range settings in the latter case (? = unknown, i.e. must be provided by user or determined from data).

• [x] (maximum) homological degree: Yes; $[0, ?]$
• [ ] scale sequence(s): Yes; TBD
• [ ] number of levels (PL): TBD
• [ ] weight exponent (PS): TBD
• [ ] standard deviation (PI): TBD
• [ ] side length (PB): TBD

[corybrunson]

@uislambekov are there any parameters that you think should not be tunable? It makes sense to me to allow tuning them all. The only one i would have questioned is the scale sequence (or sequences xSeq & ySeq), but we discussed that today.