Refining Clauses Not Correctly Identified

[SamuelAppleby]

When running the mining pipeline on the following worlds, if the code related to issue #37 is commented the result is as follows:

While ChainSuccession and Exists were identified for refinement, neither CoExistence or Absence were identified or extracted for refinement.

world_0.txt
world_1.txt

Absence(A, true, 3 )
Absence(B, true, 3 )
Absence(C, true, 2 )
CoExistence(C, true, A, true )
CoExistence(C, true, B, true )
ChainSuccession(A, (A.y >= 7.5 ∧ A.x >= 1.5) ∨ (1.5 ≤ A.y ≤ 7.5 ∧ A.x >= 1.5), B, true )
Exists(A, (A.y >= 7.5 ∧ A.x >= 1.5) ∨ (1.5 ≤ A.y ≤ 7.5 ∧ A.x >= 1.5), 1 )
~~~ model ~~~
Absence(A, true, 3 )
Absence(B, true, 3 )
Absence(C, true, 2 )
CoExistence(C, true, A, true )
CoExistence(C, true, B, true )
ChainSuccession(A, (A.y <= 1.5 ∧ A.x >= 1.5) ∨ (A.y <= 7.5 ∧ 1.5 ≤ A.x ≤ 1.5), B, true )
Exists(A, (A.y <= 1.5 ∧ A.x >= 1.5) ∨ (A.y <= 7.5 ∧ 1.5 ≤ A.x ≤ 1.5), 1 )

[jackbergus]

This might be due to the fact that no sensible refinement was fined for the clauses, and therefore discarded (as below the fitness measure).

[SamuelAppleby]

This might be due to the fact that no sensible refinement was fined for the clauses, and therefore discarded (as below the fitness measure). Il giorno lun 12 giu 2023 alle ore 18:23 SamuelAppleby < @.> ha scritto:
…
Assigned #40 <#40> to @jackbergus https://github.com/jackbergus. — Reply to this email directly, view it on GitHub <#40 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAKE55T5UNNCKFYJHPZPJ23XK5GBDANCNFSM6AAAAAAZDWCBD4 . You are receiving this because you were assigned.Message ID: @.>

In that scenario, should this be detected/removed after classifier_mining, or during it?

[jackbergus]

Ok, I get your point. No, this is related to the fact that in the current pipeline, I deactivated the refinement for the following clauses:

• Absence: this needs to be handled with extreme care, and extracting no payload condition from this in a naïve way as for the other clauses will resort in correct activation conditions for an absence!
• CoExsistence and Choice, as they have two distinct activation conditions, and even this case needs to be handled with extreme care: this was partially held at the data extraction level by checking whether, when investigating the activation via DecisionTree, the payload was referring to the first argument, and when investigating the target we were referring to the second activation. Still, I did not have enough time to ascertain the correctness of the code, and I resorted to a last-minute measure to avoid refining those clauses, as such cases might be handled with extreme care.

I am referring to the following line in pattern_mining.cpp:

////////////////////////////////////////////////////////////////////////////
/// Removing the clauses having neither activation nor target conditions ///
////////////////////////////////////////////////////////////////////////////
std::vector<size_t> index;
std::unordered_set<std::string> toExclude{"Absence", "CoExistence", "Choice"}; // TODO: dealing correctly with Absence!