CpSc 421

Homework 2

Due: Sept. 26

1. (20 points +10 points extra credit) Write regular expressions that generate each of the languages below. For each language, the alphabet, , is {0, 1}. (a) A1 = {s | s contains the substring 1011}. Solution: 1011 (b) A2 = {s | |s| = 5m + 7n with m, n N }. Solution: (5 ) (7 ) , where 5 = and likewise for 7 . (c) A3 = {s | s contains an even number of 0s}. Solution: 1 (01 0) 1 (d) A4 = {s | s contains an odd number of 1s}. Solution: 0 10 (10 1) 0 (e) (10 points extra credit): A5 = {s | s contains an even number of 0s and an odd number of 1s}. Solution: Heres a DFA that recognizes the language (q0 is th initial state, q$ is the accepting state). q0 -e-> q00 q00 -0-> q01, q00 -1-> q10 q01 -0-> q00, q01 -1-> q11 q10 -0-> q11, q10 -1-> q00 q11 -0-> q10, q10 -1-> q01 q10 -e-> q$ This DFA is a GNFA and we use the procedure from the Sept. 19 lecture notes: delete q11: q0 -e-> q00 q00 -0-> q01, q01 -0-> q00, q10 -00-> q10, q10 -e-> q$

q00 -1-> q10 q01 -11-> q01, q01 -10-> q10 q10 -01-> q01, q10 -1-> q00

delete q01: q0 -e-> q00 q00 -0(11)*0-> q00, q00 -(0(11)*10 U 1)-> q10, q10 -(1 U 01(11)*0)-> q00, q10 -(00 U 01(11)*10)-> q10, q10 -e-> q$ delete q00: q0 -(0(11)*0)*(1 U 01(11)*0) -> q10, q10 -(00 U 01(11)*10 U (1 U 01(11)*0) (0(11)*0)* (1 U 0(11)*10)-> q10, -> q10, q10 -e-> q$ Let r0 = 0(11)*0

r1 = (1 U 01(11)*0) Then, q0 q10 q10 weve got -(r0*r1)-> q10, -(r0 U r1(r0*)r1)-> q10, -e-> q$

Now, we eliminate q10 to get q0 -((r0*r1)(r0 U r1(r0*)r1)*)-> q$ Thus, our solution is: (r0*r1) (r0 U r1(r0*)r1)* with r0 = 0(11)*0 and r1 = (1 U 01(11)*0) as defined above. Ill work on prettier typesetting later. 2. (30 points) In the problems below, let R1 , R2 , . . . be arbitrary regular expressions over an arbitrary nite alphabet. For each proposed identity, either prove it, or give a counter-example. Two are valid identities for which a correct proof is worth 10 points; two are not valid identities for which a counter-example is worth 5 points. (a) R1 = R1 . Solution: False. Let R1 = 0. Then R1 but / R1 , and therefore R1 = R1 .
(b) R1 R1 = R1 R1 . Solution: } = {xy |x R1 , y R1 True. Using the denitions of concatenation and Kleen-star, R1 R1 and R1 = {x1 x2 . . . xk |k 0, xi R1 i} therefore, R1 R1

= = = = =

{xx1 . . . xk |x R1 , k 0, xi R1 i} {x1 x2 . . . xk+1 |k 0, xi R1 i} {x1 . . . xk x|x R1 , k 0, xi R1 i} {yx|x R1 , y R1 } R1 R1

(c) R1 (R2 R3 ) = (R1 R2 ) (R1 R3 ). Solution: True. R1 (R2 R3 ) = {xy |x R1 , y (R2 R3 )} = {xy |x R1 , y R2 y R3 }. (R1 R2 ) (R1 R3 ) = {x|x (R1 R2 ) x (R1 R3 )} = {xy |(x R1 y R2 ) (x R1 y R3 )} = {xy |x R1 , y R2 y R3 }. (d) R1 (R2 R3 ) = (R1 R2 ) (R1 R3 ). Solution: False. Let R1 = R2 = R3 = 0. Then R1 (R2 R3 ) = {0, 00} = {00} = (R1 R2 ) (R1 R3 ). 3. (20 points) For any language, A, let AR = {s | sR A} where sR is the reverse of s as dened in homework 0: R (x c)R Prove that if A is regular, then so is AR . = = c xR

Solution: Construct an NFA for AR . Because A is regular, we can represent it with an DFA. If we reverse the arcs between states and swap the start and accepting states, we get an NFA that recognizes AR . In the stuff that follows, Ill formalize this description, take care of a few technical details, and then prove that it works as advertised. A is an regular language. Let M = (Q, , , q0 , F ) be an DFA such that L(M ) = A. Choose qx such that qx Q (i.e. qx is a new state), and let N = (Q {qx }, , R , qx , {q0 }), where R (q, c) = R (qx , ) = R (qx , c) = {p| (p, c) = q }, reverse the arcs, q = qx F, start with an move to a nal state of M , force that initial move

Ill now prove that L(N ) = AR . Because N is an NFA, L(N ) is regular. Thus, this will show that AR is regular. The key to the proof is that after reading some string, wR , the set of possible states of N are exactly those states from which M could read w and reach an accepting state. The proof is by induction on w. Induction Hypothesis: p ( R ({qx }, wR ) Q) (p, w) F . Base case, w = : p R ({qx }, wR ) Q p R ({qx }, R ) Q, p R ({qx }, ) Q, p ({qx } F ) Q, (p F ) (p, ) = F, w= = R For any set, B , R (B, ) = B (F Q) (qx Q) For any state, q , (q, ) = q

I showed all of the steps for completeness. It would be sufcient to write: p R ({qx }, ) Q pF (p, ) F Induction step, w = c x: Noting that (c x)R = xR c, we need to prove p R ({qx }, xR c) r R ({qx }, xR ). p R (r, c), r R ({qx }, xR ). (p, c) = r, (p, c) R ({qx }, xR ) ( (p, c), x) F, (p, c x) F, def. R for strings def. R for symbols induction hypothesis def. for strings

Intuitively, what this argument says is that if N can reach some state, p, by reading xR c; then it did it by rst reaching some state, r, by reading xR , and then got to state p by reading c. We then take advantage that R is the reversal of . Thus, M will go from p to r by reading c. Finally, we use the induction hypothesis with r and x to conclude that M will go from r to some state in F by reading x. I will accept an intuitive argument like this one, or the mathematical version that I stated rst. 4. (40 points): For each language below, determine whether or not the language is regular. If it is regular, draw a DFA that accepts it and write a short explanation of how your DFA works. If it is not regular, provide a proof. For each language, the alphabet, , is {0, 1}. The notation #0(s) refers to the number of #0s in s, and #1(s) refers to the number of #1s.

Figure 1: The 4 states represent the parity of the number of 0s and the parity of the number of 1s seen so far: (even, even); (odd, even); (even, odd); (odd, odd).

(a) B1 = {s | s contains an even number of 0s and an odd number of 1s}. Solution: (see DFA in Figure 1) B1 is regular: (b) B2 = {s | #1(s) = k #0(s) for some k N}. Solution: B2 is not regular: Let p be a proposed pumping lemma constant, and let w = 0p 1p B2 . For any xyz = w with |xy | p, xy 2 z = 0p+|y| 1p has more 0s than 1s, and therefore is not in B2 . It follows by the pumping lemma that B2 is not regular. (c) B3 = {s | (|#0(s) #1(s)| mod 3) = 0 }. Solution: (see DFA in Figure 2) B3 is regular:

Figure 2: The 3 states track the value of the number of 0s seen thus far minus the number of 1s seen so far mod 3. (d) B4 = {s | (|#0(s) #1(s)| mod 3) = 1 }. Solution: B4 is not regular: Let p be a proposed pumping lemma constant, and let w = 0p 1p+1 B4 . For any xyz = w with |xy | p, we consider three cases:

 k1 3k, 3k + 1, k 0 |y | = 3k + 2, k 0 If |y | = 3k , then xy 2 z = w has #0(w) = p + 3k and #1(w) = p + 1, so |#0(w) #1(w)| (mod 3) = |p + 3k (p + 1)| (mod 3) = 2 and w / B4 . If |y | = 3k + 1, then xy 2 z = w has #0(w) = p + 3k + 1 and #1(w) = p + 1, so |#0(w) #1(w)| (mod 3) = |p + 3k + 1 (p + 1)| (mod 3) = 0 and w / B4 . If |y | = 3k + 2, then xy 3 z = w has #0(w) = p + 6k + 4 and #1(w) = p + 1, so |#0(w) #1(w)| (mod 3) = |p + 6k + 4 (p + 1)| (mod 3) = 0 and w / B4 . Therefore, B4 is not regular by the pumping lemma.