4 \begin{frame
}{Semantic Branching
}
6 \item Idea: Add additional information when considering second
7 alternative of a disjunction: $
{\phi \vee \psi}$ becomes $
{\phi
8 \vee (¬
\phi \wedge \psi)
}$
9 \item Do not re-explore branches already known to be unsatisfiable already
13 \begin{frame
}{Non-atomic Negations
}
15 \item Consider $
\heartsuit\heartsuit\heartsuit{}p
\wedge \overline{\heartsuit}\overline{\heartsuit}\overline{\heartsuit}¬p$
16 \item Same as $
\heartsuit\heartsuit\heartsuit{}p
\wedge
17 ¬
\heartsuit\heartsuit\heartsuit{}p$ and thus obviously unsatisfiable
18 \item COOL normalizes formulas to make best use of Global Caching
22 \begin{frame
}{Subset Nodes
}
24 \item Lemma: If some Node is already satisfiable, all Nodes with
25 subsets of formulas, deferrals are also satisfiable
26 \item Too expensive to check in general
\pause
27 \item Might be interesting for Nodes with full focus and its
28 siblings with the same set of formulas
29 \item Cheap but does this happen often?
33 \begin{frame
}{Normalization
}
36 \item Global Caching better with more same formulas
37 \item Preprocessing in COOL currently does not consider
38 (propositional) Comutativity, Associativity or Distributivity laws
42 \begin{frame
}{Removing Edges
}
43 \begin{block
}{Propagation
}
45 \item Runs in $
\mathcal{O
}(
\mathit{|edge|
}\cdot\mathit{|nodes|
})$
46 \item Needs to consider all satisfiable nodes
47 \item Iterates over all edges
51 \item Idea: Remove unneeded edges
52 \item Cores are $
\exists$-quantified -- all unsatisfiable children
54 \item States are essentially $
\forall$-quantified -- all satisfiable
55 children may be removed
62 %%% TeX-master: "../vortrag"
63 %%% TeX-engine: luatex