9. Thue-Morse Seqeunce
Definition (1)
Thue-Morse sequence is defined as
to construct magic squares.
The Thue–Morse word
that a binary word is a word over the alphabet {0, 1}.
ition 1.1. The Thue-Morse word t = t0t1t2 · · · is the bina
→ {0, 1} defined recursively by: t0 = 0; and for n ≥ 0, t2n =
= ¯tn, where ¯a = 1 − a for a ∈ {0, 1}. (See Figure 1.1.)
t = t0 t1 t2 t3 · · · tm · · · t2m t2m+1 · · ·
= 0 1 1 0 · · · a · · · a ¯a · · · .
Figure 1.1: The Thue-Morse word t. Here a ∈ {0, 1}.
10. Thue-Morse Seqeunce
Definition (1)
Thue-Morse sequence is defined as
to construct magic squares.
The Thue–Morse word
that a binary word is a word over the alphabet {0, 1}.
ition 1.1. The Thue-Morse word t = t0t1t2 · · · is the bina
→ {0, 1} defined recursively by: t0 = 0; and for n ≥ 0, t2n =
= ¯tn, where ¯a = 1 − a for a ∈ {0, 1}. (See Figure 1.1.)
t = t0 t1 t2 t3 · · · tm · · · t2m t2m+1 · · ·
= 0 1 1 0 · · · a · · · a ¯a · · · .
Figure 1.1: The Thue-Morse word t. Here a ∈ {0, 1}.
1 The Thue–Morse word
call that a binary word is a word over the alphabet {0, 1}.
finition 1.1. The Thue-Morse word t = t0t1t2 · · · is the binary
N → {0, 1} defined recursively by: t0 = 0; and for n ≥ 0, t2n =
+1 = ¯tn, where ¯a = 1 − a for a ∈ {0, 1}. (See Figure 1.1.)
t = t0 t1 t2 t3 · · · tm · · · t2m t2m+1 · · ·
= 0 1 1 0 · · · a · · · a ¯a · · · .
Figure 1.1: The Thue-Morse word t. Here a ∈ {0, 1}.
ample. Here are the first forty letters of the Thue–Morse word,
t = 0110100110010110100101100110100110010110 · · ·
12. Definition (3)
For every , let denote the sum
of the digits in the binary expansion of .
Then, Thue-Morse sequence is defined as
Thue-Morse Seqeunce
ホワイトボードで!
16. Sequenceにおける「繰り返し」
Definition
A square is a string of the form for some string .
A word is square-free if it contains no subword that is
square.
An overlap is a string of the form for some
string and some single letter .
17. Sequenceにおける「繰り返し」
Definition
A square is a string of the form for some string .
A word is square-free if it contains no subword that is
square.
An overlap is a string of the form for some
string and some single letter .
A word is overlap-free if it contains no subword that is
overlap.
18. Sequenceにおける「繰り返し」
Definition
A square is a string of the form for some string .
A word is square-free if it contains no subword that is
square.
An overlap is a string of the form for some
string and some single letter .
A word is overlap-free if it contains no subword that is
overlap.
(証明は省略.結構ややこしい)
Thue-Morse sequence is overlap-free.
Fact
24. Question
Does there exists a square-free sequence
over the alphabet
Axel Thue (1863–1922)
5 / 55
Axel Thue
存在する.
Thue-Morse sequence を
使って構成できる.
Avoidability in words という分野の芽吹き
25. ホワイトボードで!
Theorem (Thue)
where sn = (−1)tn
.
P = (∗), Q =
∞
n≥0
2n
2n + 1
sn
For n ≥ 1, define cn to be the number of 1’s between the n-th
and (n + 1)-th occurence of 0 in the Thue-Morse sequence t.
Then the seqeunce c = 210201 · · · is a square-free sequence
over the alphabet Σ3.
Avoidability in words という分野の芽吹き
26. Avoidability in words という分野の芽吹き
最近でも色々activeに研究されてるらしい.
詳しくはShallit先生による解説スライド
“The Ubiquitous Thue-Morse Sequence” を
チェック!
https://cs.uwaterloo.ca/~shallit/Talks/green3.pdf
27. おまけ: Thue-Morse sequenceと級数
CHAPTER 1. THE THUE–MORSE W
f. Note that d2 satisfies the following recurrence relations: d2(0)
n) = d2(n); and d2(2n + 1) = d2(n) + 1. Since d2(n) mod 2 satisfie
recurrences defining tn, we have tn = d2(n) mod 2.
rcise 1.1. If t = t0t1t2 · · · is the Thue-Morse word, show that
n≥0
(−1)tn
xn
= (1 − x)(1 − x2
)(1 − x4
)(1 − x8
) · · · .
rcise 1.2 ([AS1999]). Let t = t0t1t2 · · · be the Thue-Morse word
n = (−1)tn
for n ≥ 0. Compute the following.
1
2
s0
3
4
s1
5
6
s2
· · ·
2i + 1
2i + 2
si
· · · .
Definition(3)から成り立つことが自明.
37. Can an infinite game of chess occur under this
The question was answered by Max Euwe, the
(and world champion from 1935–1937) in 192
Figure: Max Euwe (1901–19
可能だよ!
Thue-Morse sequence を
使って構成できるよ!
Max Euwe
(1901-1981)
おまけ: Thue-Morse sequenceとチェス
1935-1937 年度
チェス世界王者
千日手ルールを「同手順が3回
連続したら」に緩めた場合は?
38. おまけ: Thue-Morse sequenceとチェス
= 0 1 1 0 · · · a · · · a ¯a · · · .
Figure 1.1: The Thue-Morse word t. Here a ∈ {0, 1}.
mple. Here are the first forty letters of the Thue–Morse word,
t = 0110100110010110100101100110100110010110 · · ·
Our first characterization of the Thue-Morse word is in terms of b
ansions of nonnegative integers. For every n ∈ N, let d2(n) denot
of the digits in the binary expansion of n.
position 1.2. For all n ∈ N, we have tn = d2(n) mod 2.
83
Thue-Morse sequence
に対して, の時
OTHER SEMI-OPEN GAM
They start:
1. e2-e4
XABCDEFGH
8rsnlwqkvlntr(
7zppzppzppzpp'
6-+-+-+-+&
5+-+-+-+-%
4-+-+P+-+$
3+-+-+-+-#
2PzPPzP-zPPzP"
1tRNvLQmKLsNR!
Xabcdefgh
WHITE SAYS:
Nb1-c3 Nb8-c6 Nc3-b1 Nc6-b8
の時
Ng1-f3 Ng8-f6 Nf3-g1 Nf6-g8
と進めれば同一手順を
3回以上繰り返さずに
無限手数ゲームが可能!
39. おまけ: Thue-Morse sequenceとチェス
= 0 1 1 0 · · · a · · · a ¯a · · · .
Figure 1.1: The Thue-Morse word t. Here a ∈ {0, 1}.
mple. Here are the first forty letters of the Thue–Morse word,
t = 0110100110010110100101100110100110010110 · · ·
Our first characterization of the Thue-Morse word is in terms of b
ansions of nonnegative integers. For every n ∈ N, let d2(n) denot
of the digits in the binary expansion of n.
position 1.2. For all n ∈ N, we have tn = d2(n) mod 2.
83
Thue-Morse sequence
に対して, の時
OTHER SEMI-OPEN GAM
They start:
1. e2-e4
XABCDEFGH
8rsnlwqkvlntr(
7zppzppzppzpp'
6-+-+-+-+&
5+-+-+-+-%
4-+-+P+-+$
3+-+-+-+-#
2PzPPzP-zPPzP"
1tRNvLQmKLsNR!
Xabcdefgh
WHITE SAYS:
Nb1-c3 Nb8-c6 Nc3-b1 Nc6-b8
の時
Ng1-f3 Ng8-f6 Nf3-g1 Nf6-g8
と進めれば同一手順を
3回以上繰り返さずに
無限手数ゲームが可能!
Thue-Morse sequence は
cube-free という性質を
使っている(証明略).
47. Definition
Sequenceを特徴付ける: k-morphic
A morphism is a function satisfying
for all
A morphism is prolongable on if there exists a letter
such that for some
In this case, the infinite sequence
is the unique infinite fixed point of starting with
48. Definition
Sequenceを特徴付ける: k-morphic
A morphism is a function satisfying
for all
A morphism is prolongable on if there exists a letter
such that for some
In this case, the infinite sequence
is the unique infinite fixed point of starting with
このように,ある morphism の不動点とな
る sequence を morphic sequence と呼ぶ.
49. Definition
Sequenceを特徴付ける: k-morphic
A morphism is k-uniform if for all
An infinite sequence is k-morphic if there exists a
k-uniform morphism that has as a fixed point.
Thue-Morse sequence は 2-morphic.
となるmorphismに対し,
50. Definition
Sequenceを特徴付ける: k-automatic
An infinite sequence is k-automatic if there exists a
k-DFAO such that for all the output of the automaton
when reading the word is ,with the base-k
expansion of
ホワイトボードで!
DFAOとか説明が面倒なので
Mix-Automatic Sequences 263
q0/a q1/b
0 1
1
0
ating the Thue–Morse sequence abbabaabbaababba· · ·
51. Sequenceを特徴付ける: k-automatic
Thue-Morse sequence は 2-automatic.
Mix-Automatic Sequences
q0/a q1/b
0 1
1
0
1. DFAO generating the Thue–Morse sequence abbabaabbaababba· · ·
of n. For example, for input (3)2 = 11 the automaton ends in sta
ut a, and for input (4)2 = 100 in state q1 with output b.
tomaton of Figure 1 is called a deterministic finite-state automaton
FAO). For k ≥ 2, a k-DFAO is an automaton over the input alph
0, 1, . . ., k − 1}. An infinite sequence w ∈ ∆ω
is called k-automat
52. Sequenceを特徴付ける: k-automatic
Thue-Morse sequence は 2-automatic.
Mix-Automatic Sequences
q0/a q1/b
0 1
1
0
1. DFAO generating the Thue–Morse sequence abbabaabbaababba· · ·
of n. For example, for input (3)2 = 11 the automaton ends in sta
ut a, and for input (4)2 = 100 in state q1 with output b.
tomaton of Figure 1 is called a deterministic finite-state automaton
FAO). For k ≥ 2, a k-DFAO is an automaton over the input alph
0, 1, . . ., k − 1}. An infinite sequence w ∈ ∆ω
is called k-automat
53. Sequenceを特徴付ける: k-automatic
Thue-Morse sequence は 2-automatic.
Mix-Automatic Sequences
q0/a q1/b
0 1
1
0
1. DFAO generating the Thue–Morse sequence abbabaabbaababba· · ·
of n. For example, for input (3)2 = 11 the automaton ends in sta
ut a, and for input (4)2 = 100 in state q1 with output b.
tomaton of Figure 1 is called a deterministic finite-state automaton
FAO). For k ≥ 2, a k-DFAO is an automaton over the input alph
0, 1, . . ., k − 1}. An infinite sequence w ∈ ∆ω
is called k-automat
54. Sequenceを特徴付ける: k-automatic
Thue-Morse sequence は 2-automatic.
Mix-Automatic Sequences
q0/a q1/b
0 1
1
0
1. DFAO generating the Thue–Morse sequence abbabaabbaababba· · ·
of n. For example, for input (3)2 = 11 the automaton ends in sta
ut a, and for input (4)2 = 100 in state q1 with output b.
tomaton of Figure 1 is called a deterministic finite-state automaton
FAO). For k ≥ 2, a k-DFAO is an automaton over the input alph
0, 1, . . ., k − 1}. An infinite sequence w ∈ ∆ω
is called k-automat
注意: この資料ではDFAOは常に右から左に文字列を読み進める!!
57. Automatic Sequences and Zip-Specifications
Clemens Grabmayer
Utrecht University, Dept. of Philosophy
Janskerkhof 13a, 3512 BL Utrecht, The Netherlands
Email: clemens@phil.uu.nl
J¨org Endrullis
VU University Amsterdam, Dept. of Computer Science
De Boelelaan 1081a, 1081 HV Amsterdam, The Netherlands
Email: j.endrullis@vu.nl
Dimitri Hendriks
VU University Amsterdam, Dept. of Computer Science
De Boelelaan 1081a, 1081 HV Amsterdam, The Netherlands
Email: r.d.a.hendriks@vu.nl
Jan Willem Klop
VU University Amsterdam, Dept. of Computer Science
De Boelelaan 1081a, 1081 HV Amsterdam, The Netherlands
Email: j.w.klop@vu.nl
Lawrence S. Moss
Indiana University, Dept. of Mathematics
831 East Third Street, Bloomington, IN 47405-7106 USA
Email: lsm@cs.indiana.edu
Abstract—We consider infinite sequences of symbols, also
known as streams, and the decidability question for equality of
streams defined in a restricted format. (Some formats lead to un-
decidable equivalence problems.) This restricted format consists
of prefixing a symbol at the head of a stream, of the stream
function ‘zip’, and recursion variables. Here ‘zip’ interleaves
the elements of two streams alternatingly. The celebrated Thue–
I. INTRODUCTION
Infinite sequences of symbols, also called ‘streams’, are a
playground of common interest for logic, computer science
(functional programming, formal languages, combinatorics on
infinite words), mathematics (numerations and number theory,
58. Zip-k specification
Definition
For , the function is defined by the
following rewriting rule:
Thus interleaves its argument sequences:
ホワイトボードで!
Specificationの説明が面倒なので
59. Zip specification and Thue-Morse sequence
Thue-Morse sequence は zip-2 specified
という zip-2 specification について,開始記号
MからThue-Morse sequenceが生成される.
60. Zip specification and Thue-Morse sequence
Thue-Morse sequence は zip-2 specified
という zip-2 specification について,開始記号
MからThue-Morse sequenceが生成される.
61. Zip specification and Thue-Morse sequence
Thue-Morse sequence は zip-2 specified
という zip-2 specification について,開始記号
MからThue-Morse sequenceが生成される.
62. Zip specification and Thue-Morse sequence
Thue-Morse sequence は zip-2 specified
という zip-2 specification について,開始記号
MからThue-Morse sequenceが生成される.
63. Zip specification and Thue-Morse sequence
Thue-Morse sequence は zip-2 specified
という zip-2 specification について,開始記号
MからThue-Morse sequenceが生成される.
66. zip-k specification = k-automatic
“Automatic Sequences and Zip Specifications”
(LICS’12)での成果(簡略化して紹介).
Theorem
A sequence is k-automatic if and only if is
zip-k specified.
証明にはObservation-graph なるものを使う
説明が面倒くさいのでホワイトボードで!
70. Mix-Automatic Sequences
J¨org Endrullis1
, Clemens Grabmayer2
, and Dimitri Hendriks1
1
VU University Amsterdam, The Netherlands
2
Utrecht University, The Netherlands
Abstract. Mix-automatic sequences form a proper extension of the
class of automatic sequences, and arise from a generalization of finite
state automata where the input alphabet is state-dependent. In this pa-
per we compare the class of mix-automatic sequences with the class of
72. Zip-mix specification and mix-DFAO
ホワイトボードで!
mix-DFAOとか説明が面倒なので
e state q0 has two outgoing edges, reflecting the inp
has three outgoing edges, reflecting the input alpha
q0/a q1/b
0
1
0, 1
2
Fig. 2. An example of a mix-DFAO
Numeration Systems. Clearly, the numeration sys
-DFAOs cannot be the standard base-k representat
epresentation that we let these automata operate on
mix-DFAOでは dynamic numeration system
という特殊な記数法を使う!
numeration system used for
ase-k representation. Instead,
mata operate on, the base for
digits that have already been
to the most significant digit
left). We write (n)M for the
or the automaton M. For M
of the first eight numbers are
0202 (6)M = 131202
12 (7)M = 130312
sentation) in db indicates the
example (17)M = 12022312.
264 J. Endrullis, C. Grabmayer, and D. Hendriks
Knowing the base for each digit, we can reconstru
tion as follows: 17 = 1·2·3·2+0·3·2+2·2+1 wher
the product of the bases of the lower digits. Given
the base of each of the digits is determined by the
the automaton reading the digit. The states q0 and
{0, 1} and {0, 1, 2} and thus expect the input in ba
73. zip-mix specified = mix-automatic sequence
実はここまで “Automatic Sequences and Zip
Specifications” (LICS’12)での成果.
Theorem
A sequence is mix-automatic if and only if is
zip-mix specified.
では,“Mix-Automatic Sequences” (LATA’13)
での成果は?
81. 論文で紹介されてる未解決問題
(ii) For every polynomial ϕ there is a mix-automatic sequence whose subword
complexity exceeds ϕ. As a consequence there are mix-automatic sequences
that are not morphic, since morphic sequences have quadratic subword
complexity at most.
(iii) A morphic sequence that is not mix-automatic, showing that the class of
morphic sequences is not contained in the class of mix-automatic sequences.
All of these concepts are very recent, and many interesting questions remain.
We highlight three particularly intriguing, and challenging questions:
(1) (J.-P. Allouche) Characterize the intersection of mix-automatic and morphic
sequences. (Note that at least all automatic sequences are in.)
(2) Is the following problem decidable: Given two mix-DFAOs, do they generate
the same sequence?
(3) Can Cobham’s Theorem (below) be generalized to mix-automatic sequences?
Cobham’s Theorem ([3]). Let k, ≥ 2 be multiplicatively independent (i.e.,
ka
= b
, for all a, b > 0), and let w ∈ ∆ω
be both k- and -automatic. Then w is
ultimately periodic.
In order to generalize this theorem to mix-automatic sequences, one could look
for a suitable notion of multiplicative independence for base determiners. Recall
that base determiners are themselves finite automata with output.