population protocols & ehrenfest urns...
TRANSCRIPT
![Page 1: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/1.jpg)
Population protocols& Ehrenfest urns scheme
O. Bournez, P. Chassaing, J. Cohen, L. Gerin & X. KoeglerLIX, Institut Elie Cartan, LORIA
Philippe Flajolet’s Birthday Meeting2008/12/01
![Page 2: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/2.jpg)
Population protocols :the example of the majority
Aim. Initially, to compute predicates.
![Page 3: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/3.jpg)
Population protocols :the example of the majority
Aim. Initially, to compute predicates.
Example. The majority problem: is “there is a majority of blue balls” true or false ?
![Page 4: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/4.jpg)
Population protocols :the example of the majority
Aim. Initially, to compute predicates.
Example. The majority problem: is “there is a majority of blue balls” true or false ?
![Page 5: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/5.jpg)
Population protocols :the example of the majority
Aim. Initially, to compute predicates.
Example. The majority problem: is “there is a majority of blue balls” true or false ?
![Page 6: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/6.jpg)
Population protocols :the example of the majority
Aim. Initially, to compute predicates.
Example. The majority problem: is “there is a majority of blue balls” true or false ?
Theorem (Angluin et al., 2007). A predicate is computable in the basic population protocol model if and only if it is semilinear.
![Page 7: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/7.jpg)
Population protocols :a variant
3
Aim: to compute a number p, i.e. to converge to a given proportion p of blue balls.
![Page 8: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/8.jpg)
Population protocols :a variant
3
Aim: to compute a number p, i.e. to converge to a given proportion p of blue balls.
Example. The protocol pictured on the right computes p=2-√2=0.58... . More precisely,
![Page 9: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/9.jpg)
Population protocols :a variant
3
Aim: to compute a number p, i.e. to converge to a given proportion p of blue balls.
Example. The protocol pictured on the right computes p=2-√2=0.58... . More precisely,
Theorem (Ph. Ch., Bournez et al.). The stationary distribution of the proportion of blue balls converges to p, when the number N of balls goes to ∞. The error term is O(N-1/2).
![Page 10: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/10.jpg)
Ehrenfest: the two-chambers urn
4
An urn contains N (here N=9) balls in its 2 chambers.Pick a ball at random and move it from its chamber to the other chamber, and repeat.
![Page 11: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/11.jpg)
Ehrenfest: the two-chambers urn
4
An urn contains N (here N=9) balls in its 2 chambers.Pick a ball at random and move it from its chamber to the other chamber, and repeat.
If the state Xn (here Xn=4) of the system is the number of balls on the left after n steps, then X=(Xn)n≥0 is a Markov chain with transitions defined by
![Page 12: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/12.jpg)
Ehrenfest: the two-chambers urn
4
An urn contains N (here N=9) balls in its 2 chambers.Pick a ball at random and move it from its chamber to the other chamber, and repeat.
If the state Xn (here Xn=4) of the system is the number of balls on the left after n steps, then X=(Xn)n≥0 is a Markov chain with transitions defined by
![Page 13: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/13.jpg)
Ehrenfest: the hypercube
5
![Page 14: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/14.jpg)
Ehrenfest: the hypercube
5
Each configuration can be encoded by a vector x=(x1,x2, ... ,xN) ∈ {0,1}N, in which
![Page 15: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/15.jpg)
Ehrenfest: the hypercube
5
Each configuration can be encoded by a vector x=(x1,x2, ... ,xN) ∈ {0,1}N, in which
xi = 1{ball n°i is in the left chamber} .
![Page 16: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/16.jpg)
Ehrenfest: the hypercube
5
Each configuration can be encoded by a vector x=(x1,x2, ... ,xN) ∈ {0,1}N, in which
xi = 1{ball n°i is in the left chamber} .Moving ball n°i amounts to change entry n°i of x.
![Page 17: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/17.jpg)
Ehrenfest: the hypercube
5
Each configuration can be encoded by a vector x=(x1,x2, ... ,xN) ∈ {0,1}N, in which
xi = 1{ball n°i is in the left chamber} .Moving ball n°i amounts to change entry n°i of x.Moving a random ball, amounts to jump from vertex x, along a random edge of the hypercube H={0,1}N.
![Page 18: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/18.jpg)
Ehrenfest: the hypercube
5
Each configuration can be encoded by a vector x=(x1,x2, ... ,xN) ∈ {0,1}N, in which
xi = 1{ball n°i is in the left chamber} .Moving ball n°i amounts to change entry n°i of x.Moving a random ball, amounts to jump from vertex x, along a random edge of the hypercube H={0,1}N.Since H is a regular graph, the uniform distribution is the stationary distribution for this random walk.
![Page 19: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/19.jpg)
Ehrenfest: the hypercube
5
Each configuration can be encoded by a vector x=(x1,x2, ... ,xN) ∈ {0,1}N, in which
xi = 1{ball n°i is in the left chamber} .Moving ball n°i amounts to change entry n°i of x.Moving a random ball, amounts to jump from vertex x, along a random edge of the hypercube H={0,1}N.Since H is a regular graph, the uniform distribution is the stationary distribution for this random walk.Thus the binomial distribution (N,0.5) is stationary for the Markov chain X.
![Page 20: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/20.jpg)
Balanced urn models
6
Urn evolution rule. A ball is chosen from the urn, with all balls present in the urn at that time having equal chances of being chosen, and its colour is inspected. If the ball chosen is orange, then α new orange balls and β new blue balls are placed into the urn; if the ball chosen is blue, then γ new orange balls and δ new blue balls are placed into the urn.
![Page 21: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/21.jpg)
Balanced urn models
6
Urn evolution rule. A ball is chosen from the urn, with all balls present in the urn at that time having equal chances of being chosen, and its colour is inspected. If the ball chosen is orange, then α new orange balls and β new blue balls are placed into the urn; if the ball chosen is blue, then γ new orange balls and δ new blue balls are placed into the urn.
![Page 22: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/22.jpg)
Balanced urn models
6
Urn evolution rule. A ball is chosen from the urn, with all balls present in the urn at that time having equal chances of being chosen, and its colour is inspected. If the ball chosen is orange, then α new orange balls and β new blue balls are placed into the urn; if the ball chosen is blue, then γ new orange balls and δ new blue balls are placed into the urn.
![Page 23: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/23.jpg)
Balanced urn models
6
Urn evolution rule. A ball is chosen from the urn, with all balls present in the urn at that time having equal chances of being chosen, and its colour is inspected. If the ball chosen is orange, then α new orange balls and β new blue balls are placed into the urn; if the ball chosen is blue, then γ new orange balls and δ new blue balls are placed into the urn.
![Page 24: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/24.jpg)
The associated differential system
7
Theorem (Flajolet et al., 2006). For a balanced urn model, the generating function of histories starting from state (a0,b0) and ending at (a,b) after n steps,
is given by
in which X and Y are solutions of with initial conditions (x0,y0).
![Page 25: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/25.jpg)
The associated differential system : the case of the Ehrenfest model
8
For the Ehrenfest model, the associated differential system is quite simple:
![Page 26: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/26.jpg)
The associated differential system : the case of the Ehrenfest model
8
For the Ehrenfest model, the associated differential system is quite simple:
and leads to the following equations:
![Page 27: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/27.jpg)
The associated differential system : the case of the Ehrenfest model
8
For the Ehrenfest model, the associated differential system is quite simple:
and leads to the following equations:
![Page 28: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/28.jpg)
The associated differential system : the case of the Ehrenfest model
8
For the Ehrenfest model, the associated differential system is quite simple:
and leads to the following equations:
![Page 29: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/29.jpg)
The rescaled Markov chain
9
Set:
![Page 30: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/30.jpg)
The rescaled Markov chain
9
Set:
Then, due to
![Page 31: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/31.jpg)
The rescaled Markov chain
9
Set:
Then, due to
![Page 32: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/32.jpg)
The rescaled Markov chain
9
Set:
Then, due to
one obtains, for t=k/N,
![Page 33: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/33.jpg)
The rescaled Markov chain
9
Set:
Then, due to
one obtains, for t=k/N,
![Page 34: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/34.jpg)
The rescaled Markov chain
9
Set:
Then, due to
one obtains, for t=k/N,
Thus, when N is large enough, Yt is approximately solution of the SDE:
![Page 35: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/35.jpg)
The rescaled Markov chain
9
Set:
Then, due to
one obtains, for t=k/N,
Thus, when N is large enough, Yt is approximately solution of the SDE:
![Page 36: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/36.jpg)
The Ornstein-Uhlenbeck process
10
is an Ornstein–Uhlenbeck process. It admits a Gaussian stationary distribution. The stationary variance is given by:
Any solution Yt of the SDE:
An Ornstein–Uhlenbeck process Yt has closed-form representation in terms of the Brownian motion B(t):
![Page 37: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/37.jpg)
Population protocols :the general case
11
![Page 38: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/38.jpg)
Population protocols :the general case
11
There exist 27 rules, described by triplets (a,b,c) in {0,1,2}3. In the example, (a,b,c)=(1,1,2).
![Page 39: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/39.jpg)
Population protocols :the general case
11
There exist 27 rules, described by triplets (a,b,c) in {0,1,2}3. In the example, (a,b,c)=(1,1,2).Let p denote the unique root in (0,1), of the polynomial P below.
![Page 40: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/40.jpg)
Population protocols :the general case
11
There exist 27 rules, described by triplets (a,b,c) in {0,1,2}3. In the example, (a,b,c)=(1,1,2).Let p denote the unique root in (0,1), of the polynomial P below.
![Page 41: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/41.jpg)
Population protocols :the general case
11
There exist 27 rules, described by triplets (a,b,c) in {0,1,2}3. In the example, (a,b,c)=(1,1,2).Let p denote the unique root in (0,1), of the polynomial P below.
Theorem (Ph. Ch., Bournez et al.). Among the 26 possible rules, one is trivial, 10 converge to 0 or 1. For the 16 other rules, the stationary distribution of the proportion of blue balls is concentrated around p, when the number N of balls goes to ∞. The error term is O(N-1/2).
![Page 42: Population protocols & Ehrenfest urns schemealgo.inria.fr/pf60/program_files/Chassaing.pdfPopulation protocols & Ehrenfest urns scheme O. Bournez, P. Chassaing, J. Cohen, L. Gerin](https://reader036.vdocuments.us/reader036/viewer/2022062416/61156a6ba63fe375fc5f3ca8/html5/thumbnails/42.jpg)
Bibliography
PhysicsÜber zwei bekannte Einwände gegen das Boltzmannsche H-Theorem, Paul Ehrenfest and Tatiana Ehrenfest, Physikalische Zeitschrift 8 (1907), no. 9, 311–314.On the theory of Brownian Motion, G.E.Uhlenbeck and L.S.Ornstein, Phys.Rev. 36:823–41, 1930Random walk and the theory of Brownian motion, Mark Kac, Am. Math. Monthly 54 (1947), 369–391.
Urn modelsPolya Urn Models, Hosam Mahmoud, 2008.Functional limit theorems for multitype branching processes and generalized Pólya urns, Svante Janson, Stochastic Processes and Applications 110 (2004), no. 2, 177–245.Limit theorems for triangular urn schemes, Svante Janson, PTRF 134 (2006), 417–452.Some exactly solvable models of urn process theory, Philippe Flajolet, Philippe Dumas, and Vincent Puyhaubert, DMTCS proc. XX, 2006, 57–116.
Population protocolsAn introduction to population protocols, James Aspnes and Eric Ruppert, Bull. EATCS, vol. 93, 106–125, 2007.On the Convergence of Population Protocols When Population Goes to Infinity, Philippe Chassaing, Olivier Bournez, Johanne Cohen, Lucas Gerin and Xavier Koegler, 2008.A simple protocol for fast robust approximate majority, Dana Angluin, James Aspnes, and David Eisenstat, in Proc. Distributed Computing, 21st International Symposium, pages 20–32, 2007