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.

2 State Identification 39<br />

PDS either, because it “merges” s1 and s2. Indeed, let k be the number of a’s<br />

applied before the first b. Ifk is even and the automaton was <strong>in</strong>itially <strong>in</strong> s1 or s2,<br />

it will still be <strong>in</strong> s1 or s2 after apply<strong>in</strong>g a k , and no <strong>in</strong>formation is ga<strong>in</strong>ed (notice<br />

that, s<strong>in</strong>ce the sequence is preset, we are not allowed to look at the output at<br />

that po<strong>in</strong>t). Then, when b is applied, 0 will be output and M4 will move to s1. At<br />

this po<strong>in</strong>t, there is no longer hope of know<strong>in</strong>g whether the mach<strong>in</strong>e was <strong>in</strong>itially<br />

<strong>in</strong> s1 or s2. Ifk is odd and the automaton was <strong>in</strong>itially <strong>in</strong> s3 or s4 then, after<br />

apply<strong>in</strong>g a k , we still end up at s1 or s2 and the previous argument aga<strong>in</strong> applies.<br />

Thus, we can conclude that M4 has no PDS.<br />

However, M4 has an ADS. One such ADS is shown <strong>in</strong> Fig. 2.5. It tells us to<br />

proceed as follows, when execut<strong>in</strong>g the test. Initially, apply a. Ifthemach<strong>in</strong>e<br />

outputs 0 then apply b, otherwise, apply a second a, followed by b. The rationale<br />

is as follows. If the mach<strong>in</strong>e outputs 0 after the <strong>in</strong>itial a, then we can deduce<br />

that it was <strong>in</strong>itially at state s1 or s2 and it has moved to s3 or s4, respectively.<br />

We can then dist<strong>in</strong>guish s3 and s4 simply by apply<strong>in</strong>g b. If, on the other hand,<br />

the mach<strong>in</strong>e output 1 after the <strong>in</strong>itial a, then we can deduce that it was <strong>in</strong>itially<br />

at s3 or s4 and has now moved to s1 or s2. In this case, we apply sequence ab to<br />

dist<strong>in</strong>guish the two latter states.<br />

s1<br />

0<br />

b<br />

a<br />

0 1<br />

1<br />

s2<br />

Fig. 2.5. One possible ADS for mach<strong>in</strong>e M4.<br />

History and results There is a lot of literature on the state identification problem<br />

and the related hom<strong>in</strong>g sequence and state verification problems (Chapters 1<br />

and 3). The earliest work <strong>in</strong> this field is the sem<strong>in</strong>al 1956 paper by Edward F.<br />

Moore [Moo56]. Then <strong>in</strong> the 60’s, many papers on these problems followed this<br />

work. These papers were ma<strong>in</strong>ly motivated by automata theory and switch<strong>in</strong>g<br />

circuits. An overview of the major results of these earlier works is <strong>in</strong> [Koh78]<br />

by Zvi Kohavi. Later on, works were also motivated by communication protocol<br />

design. Before the work of Lee and Yannakakis [LY94], the proposed algorithms<br />

for solv<strong>in</strong>g these problems take exponential time and there was no algorithm for<br />

comput<strong>in</strong>g adaptive dist<strong>in</strong>guish<strong>in</strong>g sequences with polynomial length. The algorithms<br />

proposed <strong>in</strong> these works had exponential time worst-case complexity and<br />

s3<br />

a<br />

0<br />

0<br />

b<br />

1<br />

s4

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

Saved successfully!

Ooh no, something went wrong!