NUMBER OF SYMBOL COMPARISONS IN QUICKSORT AND QUICKSELECT

Transcription

1 NUMBER OF SYMBOL COMPARISONS IN QUICKSORT AND QUICKSELECT Brigitte Vallée (CNRS and Université de Caen, France) Joint work with Julien Clément, Jim Fill and Philippe Flajolet

2 Plan of the talk. Presentation of the study Statement of the results The general model of source The main steps of the method Sketch of the proof. What is a tamed source?

3 Plan of the talk. Presentation of the study Statement of the results The general model of source The main steps of the method Sketch of the proof What is a tamed source?

4 The classical framework for sorting. The main sorting algorithms or searching algorithms e.g., QuickSort, BST-Search, InsertionSort,... deal with n (distinct) keys U 1, U 2,..., U n of the same ordered set Ω. They perform comparisons and exchanges between keys. The unit cost is the key comparison.

5 The classical framework for sorting. The main sorting algorithms or searching algorithms e.g., QuickSort, BST-Search, InsertionSort,... deal with n (distinct) keys U 1, U 2,..., U n of the same ordered set Ω. They perform comparisons and exchanges between keys. The unit cost is the key comparison. The behaviour of the algorithm (wrt to key comparisons) only depends on the relative order between the keys. It is sufficient to restrict to the case when Ω = [1..n]. The input set is then S n, with uniform probability.

6 The classical framework for sorting. The main sorting algorithms or searching algorithms e.g., QuickSort, BST-Search, InsertionSort,... deal with n (distinct) keys U 1, U 2,..., U n of the same ordered set Ω. They perform comparisons and exchanges between keys. The unit cost is the key comparison. The behaviour of the algorithm (wrt to key comparisons) only depends on the relative order between the keys. It is sufficient to restrict to the case when Ω = [1..n]. The input set is then S n, with uniform probability. Then, the analysis of all these algorithms is very well known, with respect to the number of key comparisons performed in the worst-case, or in the average case.

7 Here, realistic analysis of the two algorithms QuickSort and QuickSelect QuickSort (n, A): sorts the array A Choose a pivot; (k, A, A + ) := Partition(A); QuickSort (k 1, A ); QuickSort (n k, A + ).

8 Here, realistic analysis of the two algorithms QuickSort and QuickSelect QuickSort (n, A): sorts the array A Choose a pivot; (k, A, A + ) := Partition(A); QuickSort (k 1, A ); QuickSort (n k, A + ). QuickSelect (n, m, A): returns the value of the element of rank k in A. Choose a pivot; (k, A, A + ) := Partition(A); If m = k then QuickSelect := pivot else if m < k then QuickSelect (k 1, m, A ) else QuickSelect (n k, m k, A + );

9 Mean number K n of key comparisons Classical results for various algorithms of the class QuickMin(n) := QuickSelect(1, n) finds the minimum QuickMax(n) := QuickSelect (n, n) finds the maximum. QuickQuant α (n) := QuickSelect ( αn, n) finds the α quantile QuickMed(n) := QuickSelect ( n/2, n) finds the median QuickRand(n) := QuickSelect (m, n) for a rank m [1..n] R

10 Mean number K n of key comparisons Classical results for various algorithms of the class QuickMin(n) := QuickSelect(1, n) finds the minimum QuickMax(n) := QuickSelect (n, n) finds the maximum. QuickQuant α (n) := QuickSelect ( αn, n) finds the α quantile QuickMed(n) := QuickSelect ( n/2, n) finds the median QuickRand(n) := QuickSelect (m, n) for a rank m [1..n] R QuickSort(n) QuickMin(n) QuickMax(n) QuickRand(n) K n 2n log n K n 2n K n 2n K n 3n QuickQuant α (n) K n 2 [1 + H α ] n QuickMed(n) K n 2(1 + log 2)n H α = the entropy function = α log α + (1 α) log(1 α)

11 A more realistic framework for sorting. Keys are viewed as words. The domain Ω of keys is a subset of Σ, Σ = {the infinite words on some ordered alphabet Σ}. The words are compared [wrt the lexicographic order]. The realistic unit cost is now the symbol comparison.

12 A more realistic framework for sorting. Keys are viewed as words. The domain Ω of keys is a subset of Σ, Σ = {the infinite words on some ordered alphabet Σ}. The words are compared [wrt the lexicographic order]. The realistic unit cost is now the symbol comparison. The realistic cost of the comparison between two words A and B, A = a 1 a 2 a 3... a i... and B = b 1 b 2 b 3... b i... equals k + 1, where k is the length of their largest common prefix k := max{i; j i, a j = b j }= the coincidence

13 We are interested in this new cost for each algorithm: the number of symbol comparisons...and its mean value S n ( for n words)

14 We are interested in this new cost for each algorithm: the number of symbol comparisons...and its mean value S n ( for n words) How is S n compared to K n? That is the question... An initial question asked by Sedgewick in In order to also compare with text algorithms based on tries.

15 We are interested in this new cost for each algorithm: the number of symbol comparisons...and its mean value S n ( for n words) How is S n compared to K n? That is the question... An initial question asked by Sedgewick in In order to also compare with text algorithms based on tries. An example. Sixteen words drawn from the memoryless source p(a) = 1/3, p(b) = 2/3. We keep the prefixes of length 12. A = abbbbbaaabab B = abbbbbbaabaa C = baabbbabbbba D = bbbababbbaab E = bbabbaababbb F = abbbbbbbbabb G = bbaabbabbaba H = ababbbabbbab I = bbbaabbbbbbb J = abaabbbbaabb K = bbbabbbbbbaa L = aaaabbabaaba M = bbbaaabbbbbb N = abbbbbbabbaa O = abbabababbbb P = bbabbbaaaabb

16 A = abbbbbaaabab B = abbbbbbaabaa C = baabbbabbbba D =bbbababbbaab E = bbabbaababbb F = abbbbbbbbabb G = bbaabbabbaba H = ababbbabbbab I = bbbaabbbbbbb J = abaabbbbaabb K = bbbabbbbbbaa L = aaaabbabaaba M = bbbaaabbbbbb N = abbbbbbabbaa O = abbabababbbb P = bbabbbaaaabb

17

18

19

20

21

22

23

24

25

26

27 Plan of the talk. Presentation of the study Statement of the results The general model of source The main steps of the method Sketch of the proof What is a tamed source?

28 What is already known about the mean number of symbol-comparisons? For QuickSort(n), a first answer given by Janson and Fill ( 06), only for the classical binary source. For QuickSelect(n), this was studied by Fill and Nakama ( 07), only in the case of the binary source, only for QuickMin, QuickMax, QuickRand A quite involved form for the constants of the analysis. More than twenty pages of computation...

29 What is already known about the mean number of symbol-comparisons? For QuickSort(n), a first answer given by Janson and Fill ( 06), only for the classical binary source. For QuickSelect(n), this was studied by Fill and Nakama ( 07), only in the case of the binary source, only for QuickMin, QuickMax, QuickRand A quite involved form for the constants of the analysis. More than twenty pages of computation... Our work. We answer all these questions, in the case of a general source and a general algorihm of the class. There are precise restrictive hypotheses on the source, and we provide sufficient conditions under which these hypotheses hold. We provide a closed form for the constants of the analysis, for any source of the previous type. We use different methods, with (almost) no computation...

30 Case of QuickSort(n)[CFFV 08] Theorem 1. For any tamed source, the mean number S n of symbol comparisons used by QuickSort(n) satisfies S n 1 h S n log 2 n. and involves the entropy h S of the source S, defined as h S := lim 1 p w log p w, k k w Σ k where p w is the probability that a word begins with prefix w.

31 Case of QuickSort(n)[CFFV 08] Theorem 1. For any tamed source, the mean number S n of symbol comparisons used by QuickSort(n) satisfies S n 1 h S n log 2 n. and involves the entropy h S of the source S, defined as h S := lim 1 p w log p w, k k w Σ k where p w is the probability that a word begins with prefix w. Compared to K n 2n log n, there is an extra factor equal to 1/(2h S ) log n Compared to T n (1/h S ) n log n, there is an extra factor of log n.

32 Case of QuickMin(n), QuickMax(n), [CFFV 08] Theorem 2. For any weakly tamed source, the mean numbers of symbol comparisons used by QuickMin(n) and QuickMax(n) T ( ) n involve the constants c (ɛ) S c (ɛ) S c ( ) (+) S n and T n c (+) S n, which depend on probabilities pw and p(ɛ) w, (ɛ = ±) := X "!# 1 p(ɛ) w log 1 + pw. p w p (ɛ) w w Σ p w Here p ( ) w, p (+) w, p w are the probabilities that a word begins with a prefix w, with w = w and w < w or w > w or w = w.

33 Case of QuickRand(n) [CFFV 08] Theorem 3. For any weakly tamed source, the mean number of symbol comparisons used by QuickRand(n) (randomized wrt rank), satisfies T n c S n, with c S = X " X "!! 2!## log 1 + p(ɛ) w p(ɛ) w log 1 + pw, p w p w p w ɛ=± p (ɛ) w w Σ p 2 w Here p ( ) w, p ( ) w, p w are the probabilities that a word begins with a prefix w, with w = w and w < w or w > w or w = w.

34 Case of QuickQuant α (n) [CFFV 09] Work yet in progress Theorem 4. For any weakly tamed source, the mean number of symbol comparisons used by QuickQuant α (n) satisfies q n (α) ρ S (α) n with ρ S(α) = 2 X p w + p w log p w p (α,+) w log p (α,+) w p (α, ) w log p (α, ) w w S(α)!!# + 2 X w R(α) + 2 X w L(α) p w " 1 + p(α, ) w 1 p w "! p w 1 + p(α,+) w 1 p w log 1 pw p w (α, ) log 1 pw p (α,+) w Three parts depending on the position of probabilities p (ɛ) w wrt α: w L(α) iff p (+) w 1 α, w R(α) iff p ( ) w α w S(α) iff p ( ) w < α < 1 p (+) w, p (α, ) w = 1 α p (+) w, p (α, ) w = α p ( ) w!#.

35 The constants of the analysis for the binary source. c (ɛ) B = l 0 h B = log 2, c (+) B 1 2 l + 2 l l = c( ) B 2 l 1 k=1 = c(ɛ) B [ ( 1 k log )] k

36 The constants of the analysis for the binary source. c (ɛ) B = l 0 c B = l=0 h B = log 2, c (+) B 1 2 2l 1 2 l + 2 l 0 2 l 1 k=1 1 2 l = c( ) B 2 l 1 k=1 = c(ɛ) B [ ( 1 k log )] k [ ( k log(k + 1) k 2 log )] k

37 The constants of the analysis for the binary source. c (ɛ) B = l 0 c B = Numerically, l=0 h B = log 2, c (+) B 1 2 2l 1 2 l + 2 l 0 2 l 1 k=1 1 2 l = c( ) B 2 l 1 k=1 = c(ɛ) B [ ( 1 k log )] k [ ( k log(k + 1) k 2 log )] k c (ɛ) B = c B = To be compared to the constants of the number of key comparisons κ = 2 or κ = 3

38 Plan of the talk. Presentation of the study Statement of the results The general model of source The main steps of the method Sketch of the proof What is a tamed source?

39 The (general) model of source (I) A general source S produces infinite words on an ordered alphabet Σ. For w Σ, p w := probability that a word begins with the prefix w. Define a w := p w b w := p w w, w = w w < w, p w 0 w, w = w w w, p w 0 Then: u [0, 1], k 1, w = M k (u) Σ k such that u [a w, b w [.

40 The (general) model of source (I) A general source S produces infinite words on an ordered alphabet Σ. For w Σ, p w := probability that a word begins with the prefix w. Define a w := p w b w := p w w, w = w w < w, p w 0 w, w = w w w, p w 0 Then: u [0, 1], k 1, w = M k (u) Σ k such that u [a w, b w [. Example p 0 = 1/3, p 1 = 2/3 M 1 (1/2) = 1, M 1 (1/4) = 0

41 The (general) model of source (I) A general source S produces infinite words on an ordered alphabet Σ. For w Σ, p w := probability that a word begins with the prefix w. Define a w := p w b w := p w w, w = w w < w, p w 0 w, w = w w w, p w 0 Then: u [0, 1], k 1, w = M k (u) Σ k such that u [a w, b w [. Example p 0 = 1/3, p 1 = 2/3 M 1 (1/2) = 1, M 1 (1/4) = 0 p 00 = 1/12, p 01 = 3/12, p 10 = 1/2, p 11 = 1/6 M 2 (1/2) = 10, M 2 (1/4) = 01

42 The (general) model of source (I) A general source S produces infinite words on an ordered alphabet Σ. For w Σ, p w := probability that a word begins with the prefix w. Define a w := p w b w := p w w, w = w w < w, p w 0 w, w = w w w, p w 0 Then: u [0, 1], k 1, w = M k (u) Σ k such that u [a w, b w [. Example p 0 = 1/3, p 1 = 2/3 M 1 (1/2) = 1, M 1 (1/4) = 0 p 00 = 1/12, p 01 = 3/12, p 10 = 1/2, p 11 = 1/6 M 2 (1/2) = 10, M 2 (1/4) = 01 If p w 0 for w, the sequences (a w ) and (b w ) are adjacent They define an infinite word M(u) := lim k M k (u). Then, the source is alternatively defined by a mapping M : [0, 1] Σ.

43 The (general) model of source (II) For a prefix w Σ, the interval I = [0, 1] is divided into three intervals.

44 The (general) model of source (II) For a prefix w Σ, the interval I = [0, 1] is divided into three intervals. I w ( ) := [0, a w ] {u, M(u) < w}, I w (+) := [b w, 1] {u, M(u) > w}, I w := [a w, b w ] {u, M(u) begins with w}. The interval I w is the fundamental interval relative to the prefix w.

45 The (general) model of source (II) For a prefix w Σ, the interval I = [0, 1] is divided into three intervals. I w ( ) := [0, a w ] {u, M(u) < w}, I w (+) := [b w, 1] {u, M(u) > w}, I w := [a w, b w ] {u, M(u) begins with w}. The interval I w is the fundamental interval relative to the prefix w. The measure of each interval denoted by p w or p (±) w is the probability that M(u) begins with a prefix w with w = w and w = w or w < w or w > w. Remark: p ( ) w = a w, p (+) w = 1 b w, p w = b w a w

46 Natural instances of sources: Dynamical sources With a shift map T : I I and an encoding map τ : I Σ, the emitted word is M(x) = (τx, τt x, τt 2 x,... τt k x,...) T 2 x T x x T 3 x

47 Natural instances of sources: Dynamical sources With a shift map T : I I and an encoding map τ : I Σ, the emitted word is M(x) = (τx, τt x, τt 2 x,... τt k x,...) T 2 x T x x T 3 x A dynamical system, with Σ = {a, b, c} and a word M(x) = (c, b, a, c...).

48 Memoryless sources or Markov chains. = Dynamical sources with affine branches x x

49 The dynamical framework leads to more general sources. The curvature of branches entails correlation between symbols

50 The dynamical framework leads to more general sources. The curvature of branches entails correlation between symbols Example : the Continued Fraction source

51 Fundamental intervals and fundamental triangles x x

52 Plan of the talk. Presentation of the study Statement of the results The general model of source The main steps of the method Sketch of the proof What is a tamed source?

53 Three main steps for the analysis of the mean number S n of symbol comparisons

54 Three main steps for the analysis of the mean number S n of symbol comparisons (A) The Poisson model P Z does not deal with a fixed number n of keys. The number N of keys is now a random variable which follows a Poisson law of parameter Z. We first obtain nice expressions for S Z...

55 Three main steps for the analysis of the mean number S n of symbol comparisons (A) The Poisson model P Z does not deal with a fixed number n of keys. The number N of keys is now a random variable which follows a Poisson law of parameter Z. We first obtain nice expressions for S Z... (B) It is now possible to returm to the model where the number of keys is fixed. We obtain a nice exact formula for S n... from which it is not easy to obtain the asymptotics...

56 Three main steps for the analysis of the mean number S n of symbol comparisons (A) The Poisson model P Z does not deal with a fixed number n of keys. The number N of keys is now a random variable which follows a Poisson law of parameter Z. We first obtain nice expressions for S Z... (B) It is now possible to returm to the model where the number of keys is fixed. We obtain a nice exact formula for S n... from which it is not easy to obtain the asymptotics... (C) Then, the Rice formula provides the asymptotics of S n ( n ), as soon as the source is tamed.

57 Plan of the talk. Presentation of the study Statement of the results The general model of source The main steps of the method Sketch of the proof What is a tamed source?

58 (A) Dealing with the Poisson Model. In the P Z model, the number N of keys follows the Poisson law Z Zn Pr[N = n] = e n!, the mean number S(Z) of symbol comparisons for QuickX is S(Z) = [γ(u, t) + 1] π(u, t) du dt T where T := {(u, t), 0 u t 1} is the unit triangle γ(u, t):= coincidence between M(u) and M(t) π(u, t) du dt := Mean number of key-comparisons between M(u ) and M(t ) with u [u, u + du] and t [t dt, t] performed by QuickX

59 First Step in the Poisson model : The coincidence γ(u, t) An (easy) alternative expression for S(Z) = [γ(u, t) + 1] π(u, t) du dt T = w Σ T w π(u, t) du dt It involves the fundamental triangles and separates the rôles of the source and the algorithm. It is then sufficient to study the key probability π(u, t) of the algorithm (the second step). take its integral on each fundamental triangle (the third step)

60 Fundamental intervals and fundamental triangles x x

61 Study of the key probability π(u, t) of the algorithm (I) π(u, t) du dt := Mean number of key-comparisons between M(u ) and M(t ) with u [u, u + du] and t [t dt, t].

62 Study of the key probability π(u, t) of the algorithm (I) π(u, t) du dt := Mean number of key-comparisons between M(u ) and M(t ) with u [u, u + du] and t [t dt, t]. Case of QuickSort. M(u) and M(t) are compared iff the first pivot chosen in {M(v), v [u, t]} is M(u) or M(t)

63 Study of the key probability π(u, t) of the algorithm (I) π(u, t) du dt := Mean number of key-comparisons between M(u ) and M(t ) with u [u, u + du] and t [t dt, t]. Case of QuickSort. M(u) and M(t) are compared iff the first pivot chosen in {M(v), v [u, t]} is M(u) or M(t) QuickSort (n, A): sorts the array A Choose a pivot; (k, A, A + ) := Partition(A); QuickSort (k 1, A ); QuickSort (n k, A + ).

64 Study of the key probability π(u, t) of the algorithm (II) Case of QuickSelect

65 Study of the key probability π(u, t) of the algorithm (II) Case of QuickSelect M(u) and M(t) are compared in QuickMin iff the first pivot chosen in {M(v), v [0, t]} is M(u) or M(t)

66 Study of the key probability π(u, t) of the algorithm (II) Case of QuickSelect M(u) and M(t) are compared in QuickMin iff the first pivot chosen in {M(v), v [0, t]} is M(u) or M(t) M(u) and M(t) are compared in QuickMax iff the first pivot chosen in {M(v), v [u, 1]} is M(u) or M(t)

67 Study of the key probability π(u, t) of the algorithm (II) Case of QuickSelect M(u) and M(t) are compared in QuickMin iff the first pivot chosen in {M(v), v [0, t]} is M(u) or M(t) M(u) and M(t) are compared in QuickMax iff the first pivot chosen in {M(v), v [u, 1]} is M(u) or M(t) And for QuickQuant α? The idea is to compare QuicSelect with a dual algorithm, the QuickVal algorithm.

68 Here, we introduce another algorithm, the QuickVal algorithm, the dual algorithm of QuickSelect, QuickVal (n, a, A). : returns the rank of the element a in B = A {a} B := A {a} QV (n, a, B); QV (n, a, B). Choose a pivot in B; (k, B, B + ) := Partition(B); If a = pivot then QV := k else if a < pivot then QV := QV (k 1, a, B ) else QV := k+ QV (n k, a, B + );

69 Study of the key probability π(u, t) of the algorithm (III) QuickVal α := the algorithm where the key of interest is the word M(α)

70 Study of the key probability π(u, t) of the algorithm (III) QuickVal α := the algorithm where the key of interest is the word M(α) There are three facts The QuickVal α algorithm and the QuickQuant α algorithm are dual Their probabilistic behaviours are close

71 Study of the key probability π(u, t) of the algorithm (III) QuickVal α := the algorithm where the key of interest is the word M(α) There are three facts The QuickVal α algorithm and the QuickQuant α algorithm are dual Their probabilistic behaviours are close The QuickVal α algorithm is easy to deal with since M(u) and M(t) are compared in QuickVal α iff the first pivot chosen in {M(v), v [x, y]} is M(u) or M(t) where the interval [x, y] is the smallest interval that contains u, t and α.

72 Study of the key probability π(u, t) of the algorithm (III) QuickVal α := the algorithm where the key of interest is the word M(α) There are three facts The QuickVal α algorithm and the QuickQuant α algorithm are dual Their probabilistic behaviours are close The QuickVal α algorithm is easy to deal with since M(u) and M(t) are compared in QuickVal α iff the first pivot chosen in {M(v), v [x, y]} is M(u) or M(t) where the interval [x, y] is the smallest interval that contains u, t and α. [α, t] if u > α (case I) QuickMin [x(u, t), y(u, t)] := [u, α] if t < α (case II) QuickMax [u, t] if u < α < t (case III) QuickSort

73 . The three domains for the definition of the interval [x, y] (1, 1) (III) (I) (α, α) (II) (0, 0) [α, t] if u > α (case I) [x(u, t), y(u, t)] := [u, α] if t < α (case II) [u, t] if u < α < t (case III)

74 Study of the key probability π(u, t) of the algorithm (IV) For each algorithm, QuickSort, QuickMin, QuickMax, QuickVal α, the key probability π(u, t) of the algorithm defined as π(u, t) du dt := Mean number of key-comparisons between M(u ) and M(t ) with u [u, u + du] and t [t dt, t] equals [ ] 2 π(u, t) du dt = Zdu Zdt E 2 + N [x(u,t),y(u,t)] Here, N [x,y] is the number of words M(v) with v [x, y], It follows a Poisson law of parameter Z(y x). Then: π(u, t) = 2 Z 2 f 1 (Z(y x)) with f 1 (θ) := θ 2 [e θ 1+θ] This ends the second step in the Poisson model

75 The final expression for the mean number of symbol comparisons for each algorithm, in the Poisson model P Z S(Z) = 2 Z 2 w Σ T w f 1 (Z(y x)) du dt (1, 1) (III) (I) (0, 0) (II) (α, α) a w b w

76 (B) Return to the model where n is fixed. With the expansion of f 1, S(Z) = ( 1) k ϖ( k) Zk k!, k=2 is expressed with a series ϖ(s) of Dirichlet type, which depends both on the algorithm and the source. ϖ(s) = 2 w Σ T w (y x) (s+2) du dt

77 (B) Return to the model where n is fixed. With the expansion of f 1, S(Z) = ( 1) k ϖ( k) Zk k!, k=2 is expressed with a series ϖ(s) of Dirichlet type, which depends both on the algorithm and the source. ϖ(s) = 2 w Σ T w (y x) (s+2) du dt Since S ( n n! = [Zn ] e Z S(Z) ), there is an exact formula for S n S n = 2 n ( ) n ( 1) k ϖ( k) k k=2

78 (C) Using Rice formula As soon as ϖ(s) is weakly tamed in R(s) < σ 0 with σ 0 > 2, the residue formula transforms the sum into an integral: S n = n ( ) n ( 1) k ϖ( k) = 1 k 2iπ k=2 with 2 < d < min( 1, σ 0 ). d+i d i n! ϖ(s) s(s + 1)... (s + n) ds,

79 (C) Using Rice formula As soon as ϖ(s) is weakly tamed in R(s) < σ 0 with σ 0 > 2, the residue formula transforms the sum into an integral: S n = n ( ) n ( 1) k ϖ( k) = 1 k 2iπ k=2 with 2 < d < min( 1, σ 0 ). where d+i d i n! ϖ(s) s(s + 1)... (s + n) ds, Where are the leftmost singularities for ϖ(s)? For QuickSort: ϖ(s) = 2 Λ(s) s(s + 1) Λ(s) := w has always a singularity at s = 1. w Σ p s For all the QuickSelect algorithms, ϖ(s) is weakly tamed in R(s) < σ 0 with σ 0 > 1

80 Plan of the talk. Presentation of the study Statement of the results The general model of source The main steps of the method Sketch of the proof What is a tamed source?

81 What can be expected about Λ(s)? For any source, Λ(s) has a singularity at s = 1.

82 What can be expected about Λ(s)? For any source, Λ(s) has a singularity at s = 1. For a tamed source S, the dominant singularity of Λ(s) is located at s = 1, this is a simple pôle, whose residue equals 1/h S.

83 What can be expected about Λ(s)? For any source, Λ(s) has a singularity at s = 1. For a tamed source S, the dominant singularity of Λ(s) is located at s = 1, this is a simple pôle, whose residue equals 1/h S. In this case, there is a triple pôle at s = 1 for ϖ(s) s + 1 = 2 Λ(s) s(s + 1) 2 and ϖ(s) s h S 1 (s + 1) 3 s 1

84 For shifting the integral to the right, past... d = 1, other properties of Λ(s) are needed on Rs 1, more subtle Different behaviours of Λ(s) for Rs 1 where one can past d = 1...

85 For shifting the integral to the right, past... d = 1, other properties of Λ(s) are needed on Rs 1, more subtle Different behaviours of Λ(s) for Rs 1 where one can past d = 1... In colored domains, Λ(s) is meromorphic and of polynomial growth for s.

86 For shifting the integral to the right, past... d = 1, other properties of Λ(s) are needed on Rs 1, more subtle Different behaviours of Λ(s) for Rs 1 where one can past d = 1... In colored domains, Λ(s) is meromorphic and of polynomial growth for s. For dynamical sources, we provide sufficient conditions (of geometric or arithmetic type), under which these behaviours hold. For a memoryless source, they depend on the approximability of ratios log p i/ log p j

87 Conclusions. If the source S is tamed, then S n 1 h S n log 2 n. We are done!

88 Conclusions. If the source S is tamed, then S n 1 h S n log 2 n. We are done! QuickSort must be compared to the algorithm which uses tries for sorting n words. We have studied its mean cost T n in For a tamed source, T n 1 h S n log n.

89 Conclusions. If the source S is tamed, then S n 1 h S n log 2 n. We are done! QuickSort must be compared to the algorithm which uses tries for sorting n words. We have studied its mean cost T n in For a tamed source, T n 1 h S n log n. Our methods apply to all the QuickSelect algorithms, and the hypotheses needed for the source are different (less strong). They are related to properties of Λ(s) for Rs < 1.

90 Conclusions. If the source S is tamed, then S n 1 h S n log 2 n. We are done! QuickSort must be compared to the algorithm which uses tries for sorting n words. We have studied its mean cost T n in For a tamed source, T n 1 h S n log n. Our methods apply to all the QuickSelect algorithms, and the hypotheses needed for the source are different (less strong). They are related to properties of Λ(s) for Rs < 1. It is easy to adapt our results to the intermittent sources, which emits long sequences of the same symbols. In this case, S n = Θ(n log 3 n), T n = Θ(n log 2 n).

91 Open problems. What about the distribution of the average search cost in a BST? Is it asymptotically normal? We know that is the case if one counts the number of key comparisons. We already know that, for a tamed source, the average depth of a trie is asymptotically normal (Cesaratto-V, 07).

92 Open problems. What about the distribution of the average search cost in a BST? Is it asymptotically normal? We know that is the case if one counts the number of key comparisons. We already know that, for a tamed source, the average depth of a trie is asymptotically normal (Cesaratto-V, 07). Long term research projects... Revisit the complexity results of the main classical algorithms, and take into account the number of symbol-comparisons... instead of the number of key-comparisons.

93 Open problems. What about the distribution of the average search cost in a BST? Is it asymptotically normal? We know that is the case if one counts the number of key comparisons. We already know that, for a tamed source, the average depth of a trie is asymptotically normal (Cesaratto-V, 07). Long term research projects... Revisit the complexity results of the main classical algorithms, and take into account the number of symbol-comparisons... instead of the number of key-comparisons. Provide a sharp analytic classification of sources: Transfer geometric properties of sources into analytical properties of Λ(s)