The Subtree Size Profile of Plane-oriented Recursive Trees

Transcription

1 The Subtree Size Profile of Plane-oriented Recursive Trees Michael Fuchs Department of Applied Mathematics National Chiao Tung University Hsinchu, Taiwan ANALCO11, January 22nd, 2011 Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

2 Profiles of Trees Rooted tree of size n Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

3 Profiles of Trees Rooted tree of size n Node profile: # of nodes at level k Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

4 Profiles of Trees Rooted tree of size n Node profile: # of nodes at level k Subtree size profile: # of subtrees of size k. 7 8 Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

5 Profiles of Trees Rooted tree of size n Node profile: # of nodes at level k Subtree size profile: # of subtrees of size k. 7 8 Both are a double-indexed sequence X n,k. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

6 Recent Studies of Profile Node Profile: extensively studied for many classes of trees. Drmota and Gittenberger (1997); Chauvin, Drmota, Jabbour-Hattab (2001); Chauvin, Klein, Marckert, Rouault (2005); Drmota and Hwang (2005); Fuchs, Hwang, Neininger (2006); Hwang (2007); Drmota, Janson, Neininger (2008); Park, Hwang, Nicodeme, Szpankowski (2009); Drmota and Szpankowski (2010); etc. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

7 Recent Studies of Profile Node Profile: extensively studied for many classes of trees. Drmota and Gittenberger (1997); Chauvin, Drmota, Jabbour-Hattab (2001); Chauvin, Klein, Marckert, Rouault (2005); Drmota and Hwang (2005); Fuchs, Hwang, Neininger (2006); Hwang (2007); Drmota, Janson, Neininger (2008); Park, Hwang, Nicodeme, Szpankowski (2009); Drmota and Szpankowski (2010); etc. Subtree Size Profile: mainly studied for binary trees and increasing trees. Feng, Miao, Su (2006); Feng, Mahmoud, Su (2007); Feng, Mahmoud, Panholzer (2008); Fuchs (2008); Chang and Fuchs (2010); Dennert and Grübel (2010); etc. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

8 Why studying the Subtree Size Profile? Fine shape characteristic Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

9 Why studying the Subtree Size Profile? Fine shape characteristic, e.g., T n := total path length; W n := Wiener index, then, T n = kx n,k and W n = k)kx n,k. k 0 k 0(n Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

10 Why studying the Subtree Size Profile? Fine shape characteristic, e.g., T n := total path length; W n := Wiener index, then, T n = kx n,k and W n = k)kx n,k. k 0 k 0(n Contains information about occurrences of patterns. Important in many fields such as Computer Science (compressing, etc.), Mathematical Biology (phylogenetics), etc. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

11 Random Binary Trees Randomly pick an external node and replace it by a cherry. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

12 Random Binary Trees 1 Randomly pick an external node and replace it by a cherry. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

13 Random Binary Trees 1 2 Randomly pick an external node and replace it by a cherry. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

14 Random Binary Trees Randomly pick an external node and replace it by a cherry. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

15 Random Binary Trees Randomly pick an external node and replace it by a cherry. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

16 Random Binary Trees Randomly pick an external node and replace it by a cherry. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

17 Random Binary Trees Randomly pick an external node and replace it by a cherry. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

18 Random Binary Trees Randomly pick an external node and replace it by a cherry. 7 Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

19 Random Binary Trees Randomly pick an external node and replace it by a cherry. 7 8 Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

20 Random Binary Trees Randomly pick an external node and replace it by a cherry. 7 8 Equivalent to random binary search tree. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

21 Random Recursive Trees Randomly pick an external node and replace it by an internal node. Add external nodes. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

22 Random Recursive Trees 1 Randomly pick an external node and replace it by an internal node. Add external nodes. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

23 Random Recursive Trees 1 2 Randomly pick an external node and replace it by an internal node. Add external nodes. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

24 Random Recursive Trees Randomly pick an external node and replace it by an internal node. Add external nodes. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

25 Random Recursive Trees Randomly pick an external node and replace it by an internal node. Add external nodes. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

26 Random Recursive Trees Randomly pick an external node and replace it by an internal node. Add external nodes. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

27 Random Recursive Trees Randomly pick an external node and replace it by an internal node. Add external nodes. Same as uniform model for non-plane trees with increasing labels. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

28 Random Plane-oriented Recursive Trees (PORTs) Randomly pick an external node and replace it by an internal node. Add external nodes. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

29 Random Plane-oriented Recursive Trees (PORTs) 1 Randomly pick an external node and replace it by an internal node. Add external nodes. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

30 Random Plane-oriented Recursive Trees (PORTs) 1 2 Randomly pick an external node and replace it by an internal node. Add external nodes. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

31 Random Plane-oriented Recursive Trees (PORTs) Randomly pick an external node and replace it by an internal node. Add external nodes. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

32 Random Plane-oriented Recursive Trees (PORTs) Randomly pick an external node and replace it by an internal node. Add external nodes. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

33 Random Plane-oriented Recursive Trees (PORTs) Randomly pick an external node and replace it by an internal node. Add external nodes. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

34 Random Plane-oriented Recursive Trees (PORTs) Randomly pick an external node and replace it by an internal node. Add external nodes. Same as uniform model for plane trees with increasing labels. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

35 Importance of the Random Models Random Binary Trees Binary search tree, Quicksort, Yule-Harding Model in Phylogenetics, Coalescent Model, etc. Random Recursive Trees Simple model for spread of epidemics, for pyramid schemes, for stemma construction of philology, etc. Also, used in computational geometry and in Hopf algebras. Random PORTs One of the simplest network models (for instance for WWW). Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

36 Limit Laws for Random Binary Trees Theorem (Feng, Mahmoud, Panholzer; F.) (i) (Normal range) Let k = o ( n). Then, X n,k µ n,k σ n,k (ii) (Poisson range) Let k c n. Then, X n,k d N (0, 1). d Poisson(2c 2 ). (iii) (Degenerate range) Let k < n and n = o(k). Then, L X 1 n,k 0. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

37 Limit Laws for Random Binary Trees Theorem (Feng, Mahmoud, Panholzer; F.) (i) (Normal range) Let k = o ( n). Then, X n,k µ n,k σ n,k (ii) (Poisson range) Let k c n. Then, X n,k d N (0, 1). d Poisson(2c 2 ). (iii) (Degenerate range) Let k < n and n = o(k). Then, L X 1 n,k 0. Similar result holds for random recursive trees as well. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

38 Consequences for Occurrences of Pattern Sizes Pattern=Subtree on the fringe of the tree. # of patterns of size k: C k = 1 ( ) 2k 4k k + 1 k πk 3/2. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

39 Consequences for Occurrences of Pattern Sizes Pattern=Subtree on the fringe of the tree. # of patterns of size k: C k = 1 ( ) 2k 4k k + 1 k πk 3/2. Hence, all patterns can just occur up to a size O(log n). Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

40 Consequences for Occurrences of Pattern Sizes Pattern=Subtree on the fringe of the tree. # of patterns of size k: C k = 1 ( ) 2k 4k k + 1 k πk 3/2. Hence, all patterns can just occur up to a size O(log n). On the other hand, our result shows: Pattern sizes occur until o( n). Pattern sizes sporadically exist around n. Patterns with sizes beyond n are unlikely. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

41 Why are Pattern Sizes beyond n unlikely? Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

42 Why are Pattern Sizes beyond n unlikely? T n =total path length. It is well-known that T n is of order n log n with high probability. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

43 Why are Pattern Sizes beyond n unlikely? T n =total path length. It is well-known that T n is of order n log n with high probability. Recall that n 1 T n = k=0 kx n,k Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

44 Why are Pattern Sizes beyond n unlikely? T n =total path length. It is well-known that T n is of order n log n with high probability. Recall that n 1 T n = k=0 kx n,k Hence, if all pattern sizes up to k 0 exist, then Θ(k 2 0) = k k 0 k T n = Θ(n log n). Thus, pattern sizes beyond n log n are very unlikely. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

45 Method of Moments Theorem Z n, Z random variables. If E(Z m n ) E(Z m ) for all m 1 and Z is uniquely determined by its moments, then Z n d Z. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

46 Method of Moments Theorem Z n, Z random variables. If E(Z m n ) E(Z m ) for all m 1 and Z is uniquely determined by its moments, then Z n d Z. If Z is standard normal, then { E(Zn m 0, if m is odd; ) = m!/(2 m/2 (m/2)!), if m is even. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

47 Proof of the Limit Laws Feng, Mahmoud, Panholzer: Derived a complicated explicit expression for centered moments. Used the exact expression to obtain first order asymptotics. Many cancellations! Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

48 Proof of the Limit Laws Feng, Mahmoud, Panholzer: Derived a complicated explicit expression for centered moments. Used the exact expression to obtain first order asymptotics. Many cancellations! Our approach: Derived a recurrence for centered moments. Derived first order asymptotics via induction. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

49 Proof of the Limit Laws Feng, Mahmoud, Panholzer: Derived a complicated explicit expression for centered moments. Used the exact expression to obtain first order asymptotics. Many cancellations! Our approach: Derived a recurrence for centered moments. Derived first order asymptotics via induction. Our approach is easier and can be applied to other random trees. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

50 Theorem (F.) (i) (Normal range) Let k = o ( n). Then, where µ = 1/(2k 2 ) and X n,k µn σ n d N (0, 1), σ 2 = 8k2 4k 8 (2k 3)!! 2 (4k 2 1) 2 (k 1)! 2 4 k 1 k(2k + 1). (ii) (Poisson range) Let k c n. Then, X n,k d Poisson(2c 2 ). (iii) (Degenerate range) Let k < n and n = o(k). Then, L X 1 n,k 0. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

51 Proof for Fixed k Set Ā [m] k (z) = τ n E(X n,k µn) zn n! n 1 Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

52 Proof for Fixed k Set Then, where B [m] k Ā [m] k (z) = τ n E(X n,k µn) zn n! n 1 d dz Ā[m] k Ā[m] (z) = k (z) 1 2z [m] + B k (z), (z) is a function of Ā[i] (z) with i < m. k Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

53 Proof for Fixed k Set Then, where B [m] k Ā [m] k (z) = τ n E(X n,k µn) zn n! n 1 d dz Ā[m] k Ā[m] (z) = k (z) 1 2z [m] + B k (z), (z) is a function of Ā[i] (z) with i < m. k Above ODE has solution Ā [m] k (z) = 1 1 2z z 0 B [m] k (t) 1 2tdt. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

54 Proof for Fixed k Set Then, where B [m] k Ā [m] k (z) = τ n E(X n,k µn) zn n! n 1 d dz Ā[m] k Ā[m] (z) = k (z) 1 2z [m] + B k (z), (z) is a function of Ā[i] (z) with i < m. k Above ODE has solution Ā [m] k (z) = 1 1 2z z 0 B [m] k (t) 1 2tdt. Asymptotics of centered moments via induction ( moment-pumping ). Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

55 Singularity Analysis Consider = {z : z < r, z 1/2, arg(z 1/2) > ϕ}, where r > 1 and 0 < ϕ < π/2. Theorem (Flajolet and Odlyzko) (i) For α C \ {0, 1, 2, } [z n ](1 2z) α 2n n α 1 Γ(α) (ii) Let f(z) be analytic in. Then, ( ) α(α 1) n f(z) = O ( (1 2z) α) [z n ]f(z) = O ( 2 n n α 1). Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

56 Closure Properties Theorem Let f(z) be analytic in with f(z) = J c j (1 2z) α j + O ( (1 2z) A), j=0 where α j, A 1. Then, z 0 f(t)dt is analytic in and (i) if A > 1, then for some explicit c z 0 f(t)dt = 1 2 J j=0 (ii) if A < 1, then as above but without c. c j α j + 1 (1 2z)α j+1 + c + O ( (1 2z) A+1) ; Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

57 Asymptotic Expansions Proposition Ā [m] k (z) is analytic in. Moreover, and Ā [2m] k (z) = Ā [2m 1] k (z) = O ((1 2z) 3/2 m) (2m)!(2m 3)!!σ2m 4 m (1 2z) 1/2 m + O ( (1 2z) 1 m). m! Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

58 Asymptotic Expansions Proposition Ā [m] k (z) is analytic in. Moreover, and Ā [2m] k (z) = Ā [2m 1] k (z) = O ((1 2z) 3/2 m) (2m)!(2m 3)!!σ2m 4 m (1 2z) 1/2 m + O ( (1 2z) 1 m). m! From this, by singularity analysis, { E(X n,k µn) m 0, if m is odd; = m!/(2 m/2 (m/2)!), if m is even. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

59 Proof for Varying k Set Ā[m] n,k = E(X n,k µn) m. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

60 Proof for Varying k Set Ā[m] n,k = E(X n,k µn) m. Then, Ā [m] n,k = 1 j<n π n,j (Ā[m] j,k + Ā[m] [m] n j,k ) + B n,k, where B [m] n,k is a function of Ā[i] n,k with i < m and π n,j = 2(n j)c jc n j nc n. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

61 Proof for Varying k Set Ā[m] n,k = E(X n,k µn) m. Then, Ā [m] n,k = 1 j<n π n,j (Ā[m] j,k + Ā[m] [m] n j,k ) + B n,k, where B [m] n,k is a function of Ā[i] n,k with i < m and π n,j = 2(n j)c jc n j nc n. We have Ā [m] n,k = k+1 j n (n + 1 j)c j C n B[m] j,k. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

62 Asymptotic Expansions Proposition We have, Ā [2m 1] n,k ( ( n ) ) m 1/2 ( = o k 2, Ā [2m] n ) m n,k g m, 2k 2 where g m = (2m)!/(2 m m!). Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

63 Asymptotic Expansions Proposition Uniformly in n, k, m { Ā [m] n ( n ) }) m/2 n,k (max = O k 2, k 2. Proposition We have, Ā [2m 1] n,k ( ( n ) ) m 1/2 ( = o k 2, Ā [2m] n ) m n,k g m, 2k 2 where g m = (2m)!/(2 m m!). Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

64 Simple Classes of Increasing Trees (i) Consider rooted, plane trees with increasing labels. Let φ r be a weight sequence with φ 0 > 0 and φ r > 0 for some r 2. Denote by φ(ω) the OGF of φ r. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

65 Simple Classes of Increasing Trees (i) Consider rooted, plane trees with increasing labels. Let φ r be a weight sequence with φ 0 > 0 and φ r > 0 for some r 2. Denote by φ(ω) the OGF of φ r. Define the weight of a tree T as ω(t ) = v T φ d(v), where d(v) is the out-degree of v. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

66 Simple Classes of Increasing Trees (i) Consider rooted, plane trees with increasing labels. Let φ r be a weight sequence with φ 0 > 0 and φ r > 0 for some r 2. Denote by φ(ω) the OGF of φ r. Define the weight of a tree T as ω(t ) = v T φ d(v), where d(v) is the out-degree of v. Set τ n = ω(t ) #T =n which is the cumulative weight of all trees of size n. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

67 Simple Classes of Increasing Trees (ii) Define the probability of a tree of size n as P (T ) = ω(t ) τ n. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

68 Simple Classes of Increasing Trees (ii) Define the probability of a tree of size n as Previous Models P (T ) = ω(t ) τ n. Random binary trees: φ 0 = 1, φ 1 = 2, φ 2 = 1 and φ r = 0 for r 3; Random increasing trees: φ r = 1/r!. Random PORTs: φ r = 1. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

69 Simple Classes of Increasing Trees (ii) Define the probability of a tree of size n as Previous Models P (T ) = ω(t ) τ n. Random binary trees: φ 0 = 1, φ 1 = 2, φ 2 = 1 and φ r = 0 for r 3; Random increasing trees: φ r = 1/r!. Random PORTs: φ r = 1. All these models can be obtained from a tree evolution process. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

70 Grown Simple Classes of Increasing Trees Theorem (Panholzer and Prodinger) All simple classes of increasing tree which can be obtained via a tree evolution process are Random d-ary trees: φ(ω) = φ 0 (1 + ct/φ 0 ) d with c > 0, d {2, 3,...}; Random increasing trees: φ(ω) = φ 0 e ct/φ 0 with c > 0; Generalized random PORTs: φ(ω) = φ 0 (1 ct/φ 0 ) r 1 with c > 0, r > 1. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

71 Grown Simple Classes of Increasing Trees Theorem (Panholzer and Prodinger) All simple classes of increasing tree which can be obtained via a tree evolution process are Random d-ary trees: φ(ω) = φ 0 (1 + ct/φ 0 ) d with c > 0, d {2, 3,...}; Random increasing trees: φ(ω) = φ 0 e ct/φ 0 with c > 0; Generalized random PORTs: φ(ω) = φ 0 (1 ct/φ 0 ) r 1 with c > 0, r > 1. For stochastic properties, it is sufficient to consider the cases with φ 0 = c = 1. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

72 Mean for Grown Simple Classes of Increasing Trees Proposition For random d-ary trees, µ n,k := E(X n,k ) = d((d 1)n + 1) ((d 1)k + d)((d 1)k + 1). Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

73 Mean for Grown Simple Classes of Increasing Trees Proposition For random d-ary trees, µ n,k := E(X n,k ) = d((d 1)n + 1) ((d 1)k + d)((d 1)k + 1). Proposition For generalized random PORTs, µ n,k := E(X n,k ) = (r 1)(rn 1) (rk + r 1)(rk 1). Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

74 Limit Laws for d-ary Trees Theorem (F.) (i) (Normal range) Let k = o ( n) and k. Then, X n,k µ n,k µn,k (ii) (Poisson range) Let k c n. Then, X n,k d N (0, 1). d Poisson(2c 2 ). (iii) (Degenerate range) Let k < n and n = o(k). Then, L X 1 n,k 0. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

75 Limit Laws for Generalized Random PORTs Theorem (F.) (i) (Normal range) Let k = o ( n) and k. Then, X n,k µ n,k µn,k (ii) (Poisson range) Let k c n. Then, X n,k d N (0, 1). d Poisson(2c 2 ). (iii) (Degenerate range) Let k < n and n = o(k). Then, L X 1 n,k 0. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

76 Summary Inductive approach for deriving limit laws of subtree size profile. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

77 Summary Inductive approach for deriving limit laws of subtree size profile. Approach can be applied to many classes of random trees and shows universality of the phenomena observed by Feng, Mahmoud and Panholzer. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

78 Summary Inductive approach for deriving limit laws of subtree size profile. Approach can be applied to many classes of random trees and shows universality of the phenomena observed by Feng, Mahmoud and Panholzer. Simple classes of trees (without labels) can also be treated. For instance, random Catalan trees were considered in Chang and Fuchs (2010). Limit Laws for Patterns in Phylogenetic Trees, J. Math. Biol., 60, Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27

79 Summary Inductive approach for deriving limit laws of subtree size profile. Approach can be applied to many classes of random trees and shows universality of the phenomena observed by Feng, Mahmoud and Panholzer. Simple classes of trees (without labels) can also be treated. For instance, random Catalan trees were considered in Chang and Fuchs (2010). Limit Laws for Patterns in Phylogenetic Trees, J. Math. Biol., 60, Approach can be refined to obtain Berry-Esseen bounds, LLT and Poisson approximation results. Michael Fuchs (NCTU) Subtree Size Profile of PORTs January 22nd, / 27