07.01.2013 Views

Lecture Notes in Computer Science 3472

Lecture Notes in Computer Science 3472

Lecture Notes in Computer Science 3472

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Def<strong>in</strong>ition 9.30. Let P, Q ∈ PP.<br />

9 Test<strong>in</strong>g Theory for Probabilistic Systems 269<br />

• The trace distribution preorder � td is given by<br />

P � td Q iff tdistr(P) ⊆ tdistr(Q).<br />

• The f<strong>in</strong>ite trace distribution preorder � ftd is given by<br />

P � ftd Q iff ftdistr(P) ⊆ ftdistr(Q).<br />

• The trace distribution precongruence � tp (f<strong>in</strong>ite trace distribution<br />

precongruence � ftp , respectively) is the coarsest precongruence with respect<br />

to || 6 that is conta<strong>in</strong>ed <strong>in</strong> � td (� ftd , respectively).<br />

⊓⊔<br />

Segala shows that � td and � ftd co<strong>in</strong>cide [Seg96]. This is also stated by Stoel<strong>in</strong>ga<br />

and Vaandrager as ”Approximation Induction Pr<strong>in</strong>ciple” [SV03]. Furthermore<br />

we have that � ftd and � td characterize ⊑ may<br />

SE .<br />

Theorem 9.31. [Seg96] Let P, Q be probabilistic processes.<br />

P ⊑ may<br />

SE QiffP�tp QiffP� ftp Q.<br />

⊓⊔<br />

Segala also provides a characterization by failure distributions for ⊑must SE [Seg96].<br />

Failures are similar to traces but end <strong>in</strong> a set of actions that cannot be performed<br />

by the last state. The details of this characterization are omitted here because<br />

it is similar to the case of trace distributions.<br />

Stoel<strong>in</strong>ga and Vaandrager present an <strong>in</strong>tuitive ”test<strong>in</strong>g scenario” (also known as<br />

button push<strong>in</strong>g experiment) and proved that the result<strong>in</strong>g relation is equivalent<br />

to the trace distribution preorder [SV03]. Note that <strong>in</strong> a sense this also motivates<br />

Segala’s may-preorder due to theorem 9.31.<br />

We have also a characterization of ⊑JY by structures called ”cha<strong>in</strong>s of a process”<br />

that are similar to traces [JY02]. A very <strong>in</strong>terest<strong>in</strong>g result is the characterization<br />

by probabilistic simulation as briefly presented <strong>in</strong> the follow<strong>in</strong>g section.<br />

of ⊑ may<br />

JY<br />

9.11.3 Probabilistic Simulation<br />

The idea of ord<strong>in</strong>ary simulation is to prove that an implementation Q ref<strong>in</strong>es<br />

an abstract specification P <strong>in</strong> such a way that required properties are fulfilled<br />

[Jon91]. So Q is simulated by P if ”every step <strong>in</strong> Q can be simulated by a step<br />

<strong>in</strong> P” but not necessarily vice versa. In the probabilistic sett<strong>in</strong>g, simulation relations<br />

have been def<strong>in</strong>ed amongst others by Jonsson [JGL91] and Segala [SL94].<br />

In 2002, Jonsson and Yi proposed an alternative def<strong>in</strong>ition of probabilistic simulation<br />

which co<strong>in</strong>cides with their probabilistic may-test<strong>in</strong>g preorder [JY02].<br />

6 ArelationR is a precongruence with respect to || if P R Q implies (P || ˆ P) R (Q ||<br />

ˆP) for an arbitrary probabilistic process ˆ P.Notethat� denotes the parallel composition<br />

operator from Def<strong>in</strong>ition 9.8 here.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!