Linear constraints and manual targets with target sense equivalence

[jflowernet]

Linear constraints and manual targets with target sense equivalence

[jeffreyhanson]

Hi,

Yeah -- assuming that you have a min set objective -- using the add_manual_targets() and add_linear_constraints() functions like that would have exactly the same affect on the mathematical formulation of the optimization problem. The key difference is that encoding this formulation as a target (with add_manual_targets()) is that the functions for evaluating feature representation (eval_feature_representation_summary()) and target coverage (eval_target_coverage_summary()) will report additional information for "feature_column_name". If you don't have a min set objective, then these two implementations will not result in the same mathematical formulation. This is because -- for objectives other than the min set objective -- the targets are treated as soft constraints (meaning that solutions may not always meet the targets), whereas the linear constraints are always treated as hard constraints (regardless of which objective function is used).

How does that sound? Let me know if I can provide any further details?

[jflowernet]

Very clear, many thanks Jeff!

I feel like it might be useful to have some of these Q & As compiled somewhere for prioritizr users. I don't want to promise anything, but I could try and put an Rmarkdown together at some point. Just a thought!

[jeffreyhanson]

No worries!

Yeah that's a great idea! Yeah, no worries if you don't have time, but an FAQ document would be brilliant. We could include it as a vignette.