Differences between revisions 13 and 15 (spanning 2 versions)
Revision 13 as of 2018-09-13 13:23:49
Size: 1687
Editor: JendrikSeipp
Comment:
Revision 15 as of 2023-02-16 12:56:05
Size: 837
Comment:
Deletions are marked like this. Additions are marked like this.
Line 7: Line 7:
== Setting the "cost type" for landmark heuristics ==

For mostly historical reasons, specifying the cost type (see [[Doc/Evaluator]]) for landmark heuristics ({{{lmcount}}}, see [[Doc/Evaluator#Landmark-count_heuristic]]) is currently complicated. Here's the complete set of rules:

There are two relevant options:

 1. {{{lm_cost_type}}} of the landmark generation method (currently only used by {{{lm_rhw}}})
 2. {{{transform}}} of the landmark heuristic declaration ({{{lmcount}}}). Pass e.g. {{{transform=adapt_costs(one)}}} for unit costs.

The cost type of the landmark heuristic is determined by option 1. when using {{{admissible=false}}} and by option 2. when using {{{admissible=true}}}.

'''Note:''' Setting the cost type for the LAMA-FF synergy's master (see [[Doc/Evaluator#LAMA-FF_synergy_master]]) follows the same rules as the landmark heuristic. The cost type of the synergy's FF heuristic is determined by option 2.
Line 22: Line 9:
Unlike all other (current) heuristics, the landmark heuristic ({{{lmcount}}}, see [[Doc/Evaluator#Landmark-count_heuristic]]) requires a special setting {{{pref=true}}} to enable preferred operator support. Hence, to use preferred operators with that heuristic, '''two''' things must be done at the moment: Unlike all other (current) heuristics, the landmark heuristics ({{{landmark_sum}}} and {{{landmark_cost_partitioning}}}, see [[Doc/Evaluator#Landmark_sum_heuristic]] and [[Doc/Evaluator#Landmark_cost_partitioning_heuristic) require a special setting {{{pref=true}}} to enable preferred operator support. Hence, to use preferred operators with these heuristics, '''two''' things must be done at the moment:

Back to the HomePage.

Caveats for the options

Here, we document some aspects of the command-line syntax that may be unintuitive or complicated.

Using "preferred operators" with landmark heuristics

Unlike all other (current) heuristics, the landmark heuristics (landmark_sum and landmark_cost_partitioning, see Doc/Evaluator#Landmark_sum_heuristic and [[Doc/Evaluator#Landmark_cost_partitioning_heuristic) require a special setting pref=true to enable preferred operator support. Hence, to use preferred operators with these heuristics, two things must be done at the moment:

  1. Set pref=true in the heuristic specification.

  2. Include the heuristic in the list of heuristics in the preferred specification of the search engine.