Fundamentals of Computation Theory 23rd International Symposium FCT 2021 Athens Greece September 12 15 2021 Proceedings Lecture Notes in Computer Science 12867 Evripidis Bampis (Editor) - Instantly access the complete ebook with just one click
Fundamentals of Computation Theory 23rd International Symposium FCT 2021 Athens Greece September 12 15 2021 Proceedings Lecture Notes in Computer Science 12867 Evripidis Bampis (Editor) - Instantly access the complete ebook with just one click
com
OR CLICK HERE
DOWLOAD EBOOK
https://ebookmeta.com/product/organic-chemistry-9th-edition-john-
mcmurry/
ebookmeta.com
Advanced Heat Transfer 3rd Edition Greg F. Naterer
https://ebookmeta.com/product/advanced-heat-transfer-3rd-edition-greg-
f-naterer/
ebookmeta.com
https://ebookmeta.com/product/air-carrier-operations-4th-edition-mark-
j-holt/
ebookmeta.com
https://ebookmeta.com/product/rhysidod-the-celestial-chronicles-
book-2-1st-edition-xander-jade/
ebookmeta.com
https://ebookmeta.com/product/cultural-clinical-psychology-and-ptsd-
andreas-maercker/
ebookmeta.com
https://ebookmeta.com/product/monster-masseur-2-a-cozy-slice-of-life-
tale-of-harem-romance-and-isekai-fantasy-1st-edition-virgil-knightley/
ebookmeta.com
Museum Times : Changing Histories in South Africa 1st
Edition Leslie Witz
https://ebookmeta.com/product/museum-times-changing-histories-in-
south-africa-1st-edition-leslie-witz/
ebookmeta.com
Evripidis Bampis
Aris Pagourtzis (Eds.)
ARCoSS
LNCS 12867
Fundamentals
of Computation Theory
23rd International Symposium, FCT 2021
Athens, Greece, September 12–15, 2021
Proceedings
Lecture Notes in Computer Science 12867
Founding Editors
Gerhard Goos, Germany
Juris Hartmanis, USA
Fundamentals
of Computation Theory
23rd International Symposium, FCT 2021
Athens, Greece, September 12–15, 2021
Proceedings
123
Editors
Evripidis Bampis Aris Pagourtzis
Sorbonne University National Technical University of Athens
Paris, France Athens, Greece
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
Steering Committee
Bogdan Chlebus Augusta University, USA
Marek Karpinski (Chair) University of Bonn, Germany
Andrzej Lingas Lund University, Sweden
Miklos Santha CNRS and University Paris Diderot, France
Eli Upfal Brown University, USA
Program Committee
Evripidis Bampis (Co-chair) Sorbonne University, France
Petra Berenbrink University of Hamburg, Germany
Arnaud Casteigts University of Bordeaux, France
Marek Chrobak University of California, Riverside, USA
Hans van Ditmarsch CNRS and University of Lorraine, France
Thomas Erlebach University of Leicester, UK
Bruno Escoffier Sorbonne University, France
Henning Fernau University of Trier, Germany
Dimitris Fotakis National Technical University of Athens, Greece
Pierre Fraigniaud CNRS and University of Paris, France
Leszek Gasieniec University of Liverpool, UK, and Augusta University,
USA
Laurent Gourves CNRS and Paris Dauphine University, France
Giuseppe F. Italiano LUISS Guido Carli University, Italy
Ralf Klasing CNRS and University of Bordeaux, France
Alexander Kononov Sobolev Institute of Mathematics and Novosibirsk State
University, Russia
Antonin Kucera Masaryk University, Czech Republic
Dietrich Kuske Technische Universität Ilmenau, Germany
Nikos Leonardos National and Kapodistrian University of Athens,
Greece
Minming Li City University of Hong Kong, Hong Kong
Zsuzsanna Liptak University of Verona, Italy
Giorgio Lucarelli University of Lorraine, France
Vangelis Markakis Athens University of Economics and Business, Greece
Nicole Megow University of Bremen, Germany
Andrzej Murawski University of Oxford, UK
Aris Pagourtzis (Co-chair) National Technical University of Athens, Greece
Charis Papadopoulos University of Ioannina, Greece
Igor Potapov University of Liverpool, UK
Tomasz Radzik King’s College London, UK
viii Organization
Organizing Committee
Dimitris Fotakis (Co-chair) National Technical University of Athens, Greece
Nikos Leonardos National and Kapodistrian University of Athens,
Greece
Thanasis Lianeas National Technical University of Athens, Greece
Aris Pagourtzis (Co-chair) National Technical University of Athens, Greece
Additional Reviewers
Constantinos Daskalakis
Dániel Marx
Abstract. It is well known that hard algorithmic problems on graphs are easier
to solve if we are given a low-width tree composition of the input graph. For
many problems, if a tree decomposition of width k is available, algorithms with
running time of the form f(k)*poly(n) are known; that is, the problem is
fixed-parameter tractable (FPT) parameterized by the width of the given
decomposition. But what is the best possible function f(k) in such an algorithm?
In the past decade, a series of new upper and lower bounds gave us a tight
understanding of this question for particular problems. The talk will give a
survey of these results and some new developments.
The Complexity of Counting Problems
(Tutorial)
David Richerby
Abstract. Every computational decision problem (“Is there an X?”) has a nat-
ural counting variant (“How many X’s are there?”). More generally, computing
weighted sums such as integrals, expectations and partition functions in statis-
tical physics can also be seen as counting problems.
This tutorial will give an introduction to the complexity of solving counting
problems, both exactly and approximately. I will focus on variants of constraint
satisfaction problems. These are powerful enough to naturally express many
important problems, but also being restricted enough to allow their computa-
tional complexity to be classified completely and elegantly. No prior knowledge
of counting problems will be assumed.
Contents
Invited Papers
Contributed Papers
Fast Algorithms for the Rooted Triplet Distance Between Caterpillars . . . . . . 327
Jesper Jansson and Wing Lik Lee
1 Introduction
In the classical stable matching problem, a certain community consists of men
and women (all heterosexual and monogamous) where each person ranks those of
the opposite sex in accordance with his or her preferences for a marriage partner
(possibly declaring some matches as unacceptable). Our objective is to marry
off the members of the community in such a way that the established matching
is stable, i.e. such that there is no blocking pair. A man and a woman who are
not married to each other form a blocking pair if they prefer each other to their
mates.
In their seminal paper, Gale and Shapley [11] designed the men-proposing
deferred acceptance procedure, where men propose while women disposes. This
c Springer Nature Switzerland AG 2021
E. Bampis and A. Pagourtzis (Eds.): FCT 2021, LNCS 12867, pp. 3–17, 2021.
https://doi.org/10.1007/978-3-030-86593-1_1
4 H. Gimbert et al.
algorithm always outputs a matching which is stable, optimal for men and pes-
simal for women (in terms of rank of each person’s partner). By symmetry, there
also exists a women-optimal/men-pessimal stable matching. Gale and Shapley’s
original motivation was the assignment of students to colleges, a setting to which
the algorithm and results extend, and their approach was successfully imple-
mented in many matching markets; see for example [1,2,8,29].
However, there exists instances where the men-optimal and women-optimal
stable matchings are different, and even extreme cases of instances in which every
man/woman pair belongs to some stable matching. This raises the question of
which matching to choose [14,15] and of possible strategic behavior [9,10,28].
More precisely, if a woman lies about her preference list, this gives rise to new
stable matchings, where she will be no better off than she would be in the
true women-optimal matching. Thus, a woman can only gain from strategic
manipulation up to the maximum difference between her best and worst partners
in stable matchings. By symmetry, this also implies that the men proposing
deferred acceptance procedure is strategy-proof for men (as they will get their
best possible partner by telling the truth).
Fortunately, there is empirical evidence that in many instances, in practice
the stable matching is essentially unique (a phenomenon often referred to as
“core-convergence”); see for example [6,16,23,29]. One of the empirical expla-
nations for core-convergence given by Roth and Peranson in [29] is that the
preference lists are correlated: “One factor that strongly influences the size of
the set of stable matchings is the correlation of preferences among programs and
among applicants. When preferences are highly correlated (i.e., when similar pro-
grams tend to agree which are the most desirable applicants, and applicants tend
to agree which are the most desirable programs), the set of stable matchings is
small.”
Following that direction of enquiry, we study the core-convergence phe-
nomenon, in a model where preferences are stochastic. When preferences of
women are strongly correlated, Theorem 1 shows that the expected difference of
rank between each woman’s worst and best stable partner is a constant, hence
the incentives to manipulate are limited. If additionally the preferences of men
are uncorrelated, Theorem 2 shows that most women have a unique stable part-
ner, and therefore have no incentives to manipulate.
The following Theorem shows that under some assumptions every woman
gives approximately the same rank to all of her stable partners.
Theorem 1. Assume that each woman independently draws her preference list
from a regular distribution. The men’s preference lists are arbitrary. Let uk be
an upper bound on the odds that man mi+k is ranked before man mi :
P[mi+k w mi ]
∀k ≥ 1, uk = max w finds both m and m acceptable
P[mi w mi+k ]
i i+k
w,i
Theorem 1 is most relevant when the women’s preference lists are strongly
correlated, that is, when every woman’s preference list is “close” to a single
ranking m1 m2 . . . mM . This closeness is measured by the odds that in
some ranking, some man is ranked ahead of a man who, in the ranking m1
m2 . . . mM , would be k slots ahead of him.
We detail below three examples of applications, where the expected difference
of ranks between each woman’s best and worst partners is O(1), and thus her
incentives to misreport her preferences are limited.
– Identical preferences. If all women rank their acceptable partners using a mas-
ter list m1 m2 · · · mM , then all uk ’s are equal to 0. Then Theorem 1
states that each woman has a unique stable husband, a well-known result for
this type of instances.
– Preferences from identical popularities. Assume that women have popularity
preferences (Definition 2) and that each woman gives man mi popularity 2−i .
Then uk = 2−k and the expected rank difference is at most O(1).
– Preferences from correlated utilities. Assume that women have similar pref-
erences: each woman w gives man mi a score that is the sum of a com-
mon value i and an idiosyncratic value ηiw which is normally distributed
with mean 0 and variance σ 2 ; she then sorts men by increasing scores. Then
2
uk ≤ maxw,i {2 · P[ηiw − ηi+k w
> k]} ≤ 2e−(k/2σ) and the expected rank
√ 2
difference, by a short calculation, is at most 4 πσ 3 (1 + 2e4σ ) = O(1).
A stronger notion of approximate incentive compatibility is near-unicity of a
stable matching, meaning that most persons have either no or one unique stable
partner, and thus have no incentive to misreport their preferences. When does
that hold? One answer is given by Theorem 2.
Theorem 2. Assume that each woman independently draws her preference list
from a regular distribution. Let uk be an upper bound on the odds that man mi+k
is ranked before man mi :
P[mi+k w mi ]
∀k ≥ 1, uk = max w finds both m and m acceptable
P[mi w mi+k ]
i i+k
w,i
Further assume that all preferences are complete, that uk = exp(−Ω(k)), and
that men have uniformly random preferences. Then, in expectation the fraction
of persons who have multiple stable partners converges to 0.
Notice that in the three examples of Theorem 1, the sequence (uk )k≥1 is
exponentially decreasing. The assumptions of Theorem 2 are minimal in the
sense that removing one would bring us back to a case where a constant fraction
of woman have multiple stable partners.
– Preference lists of women. If we remove the assumption that uk is exponen-
tially decreasing, the conclusion no longer holds: consider a balanced market
balanced (M = W ) in which both men and women have complete uniformly
random preferences; then most women have ∼ ln N stable husbands [19,25].
Two-Sided Matching Markets with Strongly Correlated Preferences 7
– Preference lists of men. Assume that men have random preference built as
follows: starting from the ordering w1 , w2 , . . . , wM , each pair (w2i−1 , w2i ) is
swapped with probability 1/2, for all i. A symmetric definition for women’s
preferences satisfy the hypothesis of Theorem 2, with u1 = 1 and uk = 0 for
all k ≥ 2. Then there is a 1/8 probability that men m2i−1 and m2i are both
stable partners of women w2i−1 and w2i , for all i, hence a constant expected
fraction of persons with multiple stable partners.
– Incomplete preferences. Consider a market divided into groups of size 4 of
the form {m2i−1 , m2i , w2i−1 , w2i }, where a man and a woman are mutually
acceptable if they belong to the same group. Once again, with constant prob-
ability, m2i−1 and m2i are both stable partners of women w2i−1 and w2i .
Analyzing instances that are less far-fetched than in the worst case is the moti-
vation underlying the model of stochastically generated preference lists. A series
of papers [19,22,24–26] study the model where N men and N women have com-
plete uniformly random preferences. Asymptotically, and in expectation, the
total number of stable matchings is ∼ e−1 N ln N , in which a fixed woman has
∼ ln N stable husbands, where her best stable husband has rank ∼ ln N and her
worst stable husband has rank ∼ N/ ln N .
The first theoretical explanations of the “core-convergence” phenomenon
where given in [17] and [4], in variations of the standard uniform model. Immor-
lica and Mahdian [17] consider the case where men have constant size random
preferences (truncated popularity preferences). Ashlagi, Kanoria and Leshno [4],
consider slightly unbalanced matching markets (M < W ). Both articles prove
that the fraction of persons with several stable partners tends to 0 as the market
grows large. Theorem 2 and its proof incorporate ideas from those two papers.
Beyond strong “core-convergence”, where most agents have a unique stable
partner, one can bound the utility gain by manipulating a stable mechanism.
Lee [21] considers a model with random cardinal utilities, and shows that agents
receive almost the same utility in all stable matchings. Kanoria, Min and Qian
[18], and Ashlagi, Braverman, Thomas and Zhao [3] study the rank of each
person’s partner, under the men and women optimal stable matchings, as a
function of the market imbalance and the size of preference lists [18], or as a
function of each person’s (bounded) popularity [3]. Theorem 1 can be compared
with such results.
Beyond one-to-one matchings, school choice is an example of many-to-one
markets. Kojima and Pathak [20] generalize results from [17] and prove that most
schools have no incentives to manipulate. Azevedo and Leshno [5] show that large
markets converge to a unique stable matching in a model with a continuum of
students. To counter balance those findings, Biró, Hassidim, Romm and Shorer
[7], and Rheingans-Yoo [27] argue that socioeconomic status and geographic
preferences might undermine core-convergence, thus some incentives remain in
such markets.
8 H. Gimbert et al.
Theorem 2 (Adapted from [12]). Each person is either matched in all stable
matchings, or single in all stable matchings. In particular, a woman is matched
in all stable matchings if and only if she received at least one acceptable proposal
during Algorithm 1.
Two-Sided Matching Markets with Strongly Correlated Preferences 9
∀w ∈ μM (S) ∩ W, ∀m ∈ M \ S, μM (w) w m
Then in wn ’s preference list, the difference of ranks between wn ’s worst and best
stable partners is at most x + r − l − 1.
Proof. Since μM is woman-pessimal by Theorem 1, mn is the last stable husband
in wn ’s preference list. Let mj denote her best stable husband.
In wn ’s preference list, the interval from mj to mn contains men from her own
block, plus possibly some additional men. Such a man mi comes from outside
her block (l, r] and she prefers him to mn : since r is a prefix separator, we must
have i ≤ l. Thus x counts the number of men who do not belong to her block
but who in her preference list are ranked between mj and mn .
On the other hand, the number of men who belong to her block and who in
her preference list are ranked between mj and mn (inclusive) is at most r − l.
Together, the difference of ranks between wn ’s worst and best stable partners
is at most x + (r − l) − 1. See Fig. 1 for an illustration.
l = 2, r = 8 and x = 1
m1 m3 m7 m2 m8 m6 m9 m5 m10 w6 m4
mi with i ≤ l
mi with l < i ≤ r
µW (w6 ) µM (w6 ) unacceptable
mi with r < i
≤x+r−l−1 x = |{ |∃ ,
N
Finally, P[Ji < j] = =j P[Ji < | Ji ≤ ] ≥ k≥j−i exp(−uk ).
Recall from Lemma 3 that r − l − 1 + x is an upper bound on the difference of
rank of woman wn ’s worst and best stable husbands. We first bound the expected
value of the random variable x defined in Lemma 3.
12 H. Gimbert et al.
jumps to i − 1 and looks for a witness again. When there is no witness, a prefix
separator has been found, thus l is the largest prefix separator ≤ n−1. Similarly,
Algorithm 2 computes the smallest prefix separator r which is ≥ n. Thus, by
definition of blocks, (l, r] is the block containing wn .
0 l n r N
w1 wl+1 wn wr wN
Women
µM
Men
m1 ml+1 mn mr mN
Fig. 2. Computing the block containing woman wn . The vertical black edges corre-
spond to the men-optimal stable matching µM . There is a light gray arc (mj , wi ) if
j > i and woman wi prefers man mj to her partner: mj wi mi . The prefix sepa-
rators correspond to the solid red vertical lines which do not intersect any gray arc.
Algorithm 2 applies a right-to-left greedy method to find the largest prefix separator
l which is ≤ n − 1, jumping from dashed red line to dashed red line, and a similar
left-to-right greedy method again to find the smallest prefix separator r which is ≥ n.
This determines the block (l, r] containing n. (Color figure online)
Definition 6. Let X be the random variable defined as follows. Let (Δt )t≥0
denote a sequence of i.i.d.r.v.’s taking non-negative integer values with the fol-
lowing distribution:
∀δ > 0, P[Δt < δ] = exp − k≥δ kuk
Lemma 7. Given a woman wn , let (l, r] denote the block containing n. Condi-
tioning on H, l and r are integer random variable, such that r − n and n − 1 − l
are stochastically dominated by X.
Lemma 8. We have E[X] ≤ exp( k≥1 kuk ) k≥1 k 2 uk .
Proof (Proof of Theorem 1). Without loss of generality, we may assume that
N = M ≤ W and that each man is matched in the man-optimal stable matching
μM : to see that, for each man m we add a “virtual” woman w as his least favorite
acceptable partner, such that m is the only acceptable partner of w. A man is
14 H. Gimbert et al.
Theorem 2. Assume that each woman independently draws her preference list
from a regular distribution. Let uk be an upper bound on the odds that man mi+k
is ranked before man mi :
P[mi+k w mi ]
∀k ≥ 1, uk = max w finds both m and m acceptable
P[mi w mi+k ]
i i+k
w,i
Further assume that all preferences are complete, that uk = exp(−Ω(k)), and
that men have uniformly random preferences. Then, in expectation the fraction
of persons who have multiple stable partners converges to 0.
The proof first continues the analysis of blocks started in Sect. 2.4. When
uk = exp(−Ω(k)), it can be tightened with a mathematical analysis to prove
(Corollary 1) that with high probability, no block size exceeds O(log n), and that
in addition, in her preference list no woman switches the relative ordering of two
men mi and mi+Ω(log n) . The rest of the proof assumes that those properties
hold. The only remaining source of randomness comes from the preference lists
of men.
The intuition is that it is hard for man mi to have another stable partner
from his block. Because of the random uniform assumption on mi ’s preference
list, between wi and the next person from his block, his list is likely to have
some woman wj with j i. Woman wj likes mi better than her own partner,
because of the no-switching property, and mi likes her better than his putative
second stable partner, so they form a blocking pair preventing mi ’s second stable
partner. Transforming that intuition into a proof requires care because of the
need to condition on several events.
Lemma 9. Assume that women have preferences drawn from regular distribu-
tions such that uk = exp(−Ω(k)). Then, the size of each man’s block is a random
variable with an exponential tail:
Corollary 1. One can choose C = O(1) such that the probability of event K is
≥ 1 − 1/N 2 .
Proof (Proof of Theorem 2). As in the previous proof, in our analysis we condi-
tion on event H (see Definition 5), i.e. on (1) the man-optimal stable matching
μM , (2) each man’s ranking of the women he prefers to his partner in μM , and
(3) each woman’s ranking of the men who prefer her to their partner in μM . As
before, a person who is not matched in μM remains single in all stable match-
ings, hence, without loss of generality, we assume that M = W = N , and that
wi = μM (mi ) for all 1 ≤ i ≤ N .
Let Z denote the number of women with several stable partners. We show
that in expectation Z = O(ln2 N ), hence the fraction of persons with multi-
ple stable partners converges to 0. We separate the analysis of Z according to
whether event K holds. When K does not hold, we bound that number by N , so
by Corollary 1: E[Z] ≤ (1/N 2 ) × N + (1 − 1/N 2 ) × E(Z|K).
Conditioning on H and switching summations, we write:
E(Z|K) = P[H] · E(Z|K, H) = P[H] · P[wi has several stable husbands | K, H]
H i H
References
1. Abdulkadiroğlu, A., Pathak, P.A., Roth, A.E.: The New York city high school
match. Am. Econ. Rev. 95(2), 364–367 (2005)
2. Abdulkadiroğlu, A., Pathak, P.A., Roth, A.E., Sönmez, T.: The Boston Public
School match. Am. Econ. Rev. 95(2), 368–371 (2005)
3. Ashlagi, I., Braverman, M., Thomas, C., Zhao, G.: Tiered random matching mar-
kets: rank is proportional to popularity. In: 12th Innovations in Theoretical Com-
puter Science Conference (ITCS). Schloss Dagstuhl-Leibniz-Zentrum für Infor-
matik (2021)
4. Ashlagi, I., Kanoria, Y., Leshno, J.D.: Unbalanced random matching markets: the
stark effect of competition. J. Polit. Econ. 125(1), 69–98 (2017)
5. Azevedo, E.M., Leshno, J.D.: A supply and demand framework for two-sided
matching markets. J. Polit. Econ. 124(5), 1235–1268 (2016)
6. Banerjee, A., Duflo, E., Ghatak, M., Lafortune, J.: Marry for what? Caste and
mate selection in modern India. Am. Econ. J. Microecon. 5(2), 33–72 (2013)
7. Biró, P., Hassidim, A., Romm, A., Shorrer, R.I., Sóvágó, S.: Need versus merit: the
large core of college admissions markets. arXiv preprint arXiv:2010.08631 (2020)
8. Correa, J., et al.: School choice in Chile. In: Proceedings of the 2019 ACM Con-
ference on Economics and Computation, pp. 325–343 (2019)
9. Demange, G., Gale, D., Sotomayor, M.: A further note on the stable matching
problem. Discret. Appl. Math. 16(3), 217–222 (1987)
10. Dubins, L.E., Freedman, D.A.: Machiavelli and the Gale-Shapley algorithm. Am.
Math. Mon. 88(7), 485–494 (1981)
11. Gale, D., Shapley, L.S.: College admissions and the stability of marriage. Am.
Math. Mon. 69(1), 9–15 (1962)
12. Gale, D., Sotomayor, M.: Some remarks on the stable matching problem. Discret.
Appl. Math. 11(3), 223–232 (1985)
13. Gimbert, H., Mathieu, C., Mauras, S.: Two-sided matching markets with strongly
correlated preferences. arXiv preprint arXiv:1904.03890 (2019)
14. Gusfield, D.: Three fast algorithms for four problems in stable marriage. SIAM J.
Comput. 16(1), 111–128 (1987)
15. Gusfield, D., Irving, R.W.: The Stable Marriage Problem: Structure and Algo-
rithms. MIT Press, Cambridge (1989)
16. Hitsch, G.J., Hortaçsu, A., Ariely, D.: Matching and sorting in online dating. Am.
Econ. Rev. 100(1), 130–63 (2010)
17. Immorlica, N., Mahdian, M.: Incentives in large random two-sided markets. ACM
Trans. Econ. Comput. 3(3), 14 (2015)
18. Kanoria, Y., Min, S., Qian, P.: In which matching markets does the short side enjoy
an advantage? In: Proceedings of the 2021 ACM-SIAM Symposium on Discrete
Algorithms (SODA), pp. 1374–1386. SIAM (2021)
19. Knuth, D.E., Motwani, R., Pittel, B.: Stable husbands. In: Proceedings of the First
Annual ACM-SIAM Symposium on Discrete Algorithms, pp. 397–404 (1990)
20. Kojima, F., Pathak, P.A.: Incentives and stability in large two-sided matching
markets. Am. Econ. Rev. 99(3), 608–27 (2009)
21. Lee, S.: Incentive compatibility of large centralized matching markets. Rev. Econ.
Stud. 84(1), 444–463 (2016)
22. Lennon, C., Pittel, B.: On the likely number of solutions for the stable marriage
problem. Comb. Probab. Comput. 18(3), 371–421 (2009)
Two-Sided Matching Markets with Strongly Correlated Preferences 17
23. Pathak, P.A., Sönmez, T.: Leveling the playing field: sincere and sophisticated
players in the Boston mechanism. Am. Econ. Rev. 98(4), 1636–52 (2008)
24. Pittel, B.: The average number of stable matchings. SIAM J. Discret. Math. 2(4),
530–549 (1989)
25. Pittel, B.: On likely solutions of a stable marriage problem. Ann. Appl. Probab.
2, 358–401 (1992)
26. Pittel, B., Shepp, L., Veklerov, E.: On the number of fixed pairs in a random
instance of the stable marriage problem. SIAM J. Discret. Math. 21(4), 947–958
(2007)
27. Rheingans-Yoo, R., Street, J.: Large random matching markets with localized pref-
erence structures can exhibit large cores. Technical report, Mimeo (2020)
28. Roth, A.E.: The economics of matching: stability and incentives. Math. Oper. Res.
7(4), 617–628 (1982)
29. Roth, A.E., Peranson, E.: The redesign of the matching market for American physi-
cians: some engineering aspects of economic design. Am. Econ. Rev. 89(4), 748–780
(1999)
Communicating Finite State Machines
and an Extensible Toolchain
for Multiparty Session Types
1 Introduction
S!retry
B 1 2 3 4 5
S!title S?quote S!ok S?date
B?title B!quote B?ok B!date
S 1 2 3 4 5
B?retry
Fig. 1. Two dual communicating automata: the buyer and the seller
choice, ok and retry here, are distinct. Secondly, there is no mixed state (each
state has either only sending actions, or only receiving actions). Third, these two
machines have compatible traces (i.e. dual ): the Seller machine can be defined by
exchanging sending and receiving actions of the Buyer machine. Breaking one
of these conditions allows deadlock situations and breaking one of the first two
conditions makes the compatibility checking undecidable [16].
Essentially, the same characterisation is given in binary session types [19].
Consider the following session type of the Buyer.
The session type above describes the communication pattern using several
constructs. The operator ⊕title denotes an output of the title, whereas "e
denotes an input of a quote. The output choice features the two options ok and
retry and . denotes sequencing. end represents the termination of the session,
and μt is recursion. The simplicity and tractability of binary sessions come from
the notion of duality in interactions [15], which corresponds to compatibility of
CFSMs. In (1), not only the Buyer’s behaviour, but also the whole conversation
structure is already represented in this single type: the interaction pattern of the
Seller is fully given as the dual of the type in (1) (exchanging input ⊕ and out-
put &). When composing two parties, we only have to check they have mutually
dual types, and the resulting communication is guaranteed to be deadlock-free.
Multiparty Session Types and CFSMs. The notion of duality is no longer
effective in multiparty communication, where the whole conversation cannot be
reconstructed from only the behaviour of a single machine. Instead of directly
trying to decide whether the communication of a system satisfies safety (which is
undecidable in the general case), we devise a compatible, decidable condition of a
set of machines, which forces them to collaborate together. We define a complete
characterisation of global type behaviours into CFSMs: a set of CFSMs satisfy
some compatible conditions, if and only if the CFSMs can mimic the expected
behaviour of a given global type. A good global type means the global type can
only generate safe CFSMs by endpoint projection, which satisfies realisability.
A B C
B!quit C!finish A?quit C!save B?save A?finish
1 2 3 1 2 3 1 2 3
4 4 4
to Carol and continues. Otherwise Bob informs Carol to save the data, and
Alice gives the final notification to Carol to finish the protocol.
Deniélou and Yoshida [12] present a decidable notion of multiparty compatibil-
ity as a generalisation of duality of binary sessions, for a given set of (more than
two) CFSMs. The idea is that any single machine can see the rest of the machines
as a single machine, up to unobservable actions (like a τ -transition in CCS).
Therefore, we check the duality between each automaton and the rest, up to
internal communications (1-bounded executions in the terminology of CFSMs)
that the other machines will independently perform. For example, in Fig. 2, to
check the compatibility of trace AB!quit·AC!finish in Alice, we observe the dual
trace AB?quit · AC?finish from Bob and Carol, executing the internal commu-
nication between Bob and Carol: BC!save · BC?save. If this extended duality is
valid for all the machines from any 1-bounded reachable state, then they satisfy
multiparty compatibility and can characterise a well-formed global type.
Our motivation to study this general compatibility comes from the need
for using global types to develop tools for choreographic distributed test-
ing in web service software [44] and distributed monitoring for large cyber-
infrastructures [41], where local specifications are often updated independently
and one needs to refine the original global specification according to the local
updates.
The 1-boundedness and multiparty compatibility conditions are extended to
the k-bounded condition in [28] (called k-multiparty compatibility). Another flex-
ible form of safe and more asynchronous CFSMs (which do not rely on duality
or buffer bounds) is studied in [7,14] (called asynchronous subtyping). Unfor-
tunately, the asynchronous subtyping relation is undecidable, even if limited to
only two machines; currently its decidable sound algorithms are restricted to
either binary session types [3] or finite MPSTs [6].
Practically, a direct analysis based on CFSMs is computationally expensive,
even if the shapes of CFSMs are limited. For building a toolchain for practical
programming languages, we take the safe-by-construction approaches—we start
from specifying a global type, and project it to endpoint types or CFSMs, for
code generation into various programming languages, and/or other purposes.
Interestingly, multiparty compatibility also helps enlarge the well-formedness
condition of global types [23]. See Sect. 5.
A Global Type G
Project onto each
participant
LA LB LS
Local Type for A Local Type for B Local Type for S
send ) a message from (resp. to) the participant p, among the index set I. Recur-
sive types are realised by μt.G (resp. μt.L) and t, by taking a equi-recursive
view (However, we require types to be contractive, e.g. μt.t is not allowed).
We can obtain local types by projecting a global type upon a partici-
pant. Projection is defined as a partial function, since not all global types are
implementable—these types might be unable to be implemented in a type-safe
way. We say a global type is well-formed, if the projection of the global type upon
all participants are defined. Well-formed global types can be implemented by a
collection of concurrent processes, each implementing their projected local type.
Well-typed processes will enjoy the benefit of the MPST typing discipline, are
free from deadlocks or communication mismatches. Curious readers may refer
to [48] for more details.
From Local Types to Communicating Finite State Machines. A local
type describes the behaviour of a specific role in a given global type, which can be
represented by a communicating finite state machine 1 (CFSM) [2]. As shown by
Deniélou and Yoshida [12] and Neykova and Yoshida [35], there is an algorithm
to construct a CFSM that is trace-equivalent to the local type.
Relation to Scribble. Scribble [44,49] is a toolchain for implementing mul-
tiparty protocols. In particular, the syntax of the Scribble protocol description
language correlates closely to the theory of MPST. Neykova and Yoshida [35] give
a formal description of the Scribble protocol description language, known as
Featherweight Scribble, and establish a correspondence between global proto-
cols in Featherweight Scribble and global types in the MPST theory (Sect. 4).
In this section, we describe the structure of νScr and highlight the correspon-
dence to the multiparty session type theory. νScr is written in OCaml in around
8000 lines of code, implementing the core part of the Scribble language, with
various extensions to the original MPST. νScr also has a web interface (https://
nuscr.dev/), so that users can perform quick prototyping in browsers, saving the
need for installation (see Fig. 5 for a screenshot).
1
Also known as endpoint finite state machine (EFSM) [22].
Other documents randomly have
different content
Sitten muutama sana symmetrian ilmaantumisesta
eläinmaailmassa. Ankara tasapintainen symmetria ilmenee
sädekkäisissä, polyypeissä ja muissa alhaisemmissa eläinmuodoissa.
Korkeammissa eläimissä tasapintainen symmetria väistyy
viivasymmetrian tieltä. Ihmisvartalo järjestyy symmetrisesti
kohtisuoran viivan mukaan; tämä antaa sille vasemman ja oikean
puolen, joissa osat parittain vastaavat toisiaan ja kaikki viivat ovat
symmetrisessä sopusoinnussa. Minun tarvinnee tuskin viitata, kuinka
erikoisen tärkeää tällainen symmetria on inhimilliselle kauneudelle.
Sen ei suinkaan tarvitse olla matemaattisen tarkka, sellainen
yhdenmuotoisuus ja säännöllisyys, joka järkkymättömässä
kovuudessaan aivan kuin kieltää, että elävä elimistö on altis
ympäristön vaikutuksille ja on tulos taistelusta järjestyksen ja jonkun
muun periaatteen välillä, on pikemmin omiaan näyttämään
epämiellyttävältä tunteettoman virheettömyytensä nojalla kuin
miellyttämään. Sellainen kauneus on jäykkä, ja siitä tulee jotakin
elotonta, koska siinä ei näy jälkeäkään, mikä viittaisi, että elämä on
ollut taistelua alusta alkaen, ja että siellä, missä sitä on, ja missä se
on rikasta ja täydellisesti kehittynyttä, siellä sen myös täytyy ilmaista
jotakin suvun kärsimystaistelusta ja sen pyrkimyksestä yhä
korkeampaan vapauteen.
SYMMETRIA.
Siitä johtuu tuo horror, kauhu, joka voi vallata hänet, kun hän
näkee suuria, maanpintaa painavia ainejoukkoja, sellaisia kuin
Himalajan tai Alppien tunturit, joissa ei ilmene tasapainoa eikä
symmetriaa. Nuo vuorijoukkiot ovat tosin pienen pieniä verrattuina
kiertotähtemme omaan ainepaljouteen, mutta se ajatus, vaikkapa se
olisikin katselijan mielessä, ei kuitenkaan ole välittömästi käsillä
tarkastelun ensi hetkellä. Epäsymmetriset ainejoukkiot, jotka eivät
ole tasapainossa keskenään, näyttävät hänestä uhkaavan
maailmankaikkeuden pystyssäpysymistä tai ainakin olevan muistona
ajoista, jolloin käytiin ne taistelut, mistä tämä maailmanjärjestys on
tuloksena. Ja ihmisen kauhuun niitä katsellessa on oma
oikeutuksensa, sillä sellaiset ainejoukkiot, olkoot ne kuinka pieniä
hyvänsä itse maan massaan verrattuina, ovat kuitenkin osaltaan
syynä siihen, että maan painopiste on siirtynyt sinne, missä se on, ja
jos pohjoisnavan tienoille nousisi yhtä painava jäävuorijoukko,
järkkyisi nykyinen merenpinta paikaltaan ja nousisi meidän
seuduillamme ehkä satakin jalkaa.
VÄRISYMBOLIIKKA.