[PPT] - On P olya Urn Scheme with Infinitely Many Colors DEBLEENA THACKER PowerPoint Presentation

SLIDE 1

On P´

lya Urn Scheme with Infinitely Many Colors

DEBLEENA THACKER

Indian Statistical Institute, New Delhi Joint work with: ANTAR BANDYOPADHYAY, Indian Statistical Institute, New Delhi.

SLIDE 2

Genaralization of the Polya Urn scheme to infinitely many colors We introduce an urn with infinite but countably many colors/types of balls indexed by Z.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

2 / 17

SLIDE 3

Genaralization of the Polya Urn scheme to infinitely many colors We introduce an urn with infinite but countably many colors/types of balls indexed by Z. In this case, the so called “uniform” selection of balls does not make sense.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

2 / 17

SLIDE 4

Genaralization of the Polya Urn scheme to infinitely many colors We introduce an urn with infinite but countably many colors/types of balls indexed by Z. In this case, the so called “uniform” selection of balls does not make sense. The intial configuration of the urn U0 is taken to be a probability vector and can be thought to be the proportion of balls of each color/type we start with. Then P (A ball of color j is selected at the first trial | U0) = U0,j. We consider the replacement matrix R to be an infinite dimensional stochastic matrix.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

2 / 17

SLIDE 5

Genaralization of the Polya Urn scheme to infinitely many colors We introduce an urn with infinite but countably many colors/types of balls indexed by Z. In this case, the so called “uniform” selection of balls does not make sense. The intial configuration of the urn U0 is taken to be a probability vector and can be thought to be the proportion of balls of each color/type we start with. Then P (A ball of color j is selected at the first trial | U0) = U0,j. We consider the replacement matrix R to be an infinite dimensional stochastic matrix. At each step n ≥ 1, the same procedure as that of Polya Urn Scheme is repeated.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

2 / 17

SLIDE 6

Genaralization of the Polya Urn scheme to infinitely many colors We introduce an urn with infinite but countably many colors/types of balls indexed by Z. In this case, the so called “uniform” selection of balls does not make sense. The intial configuration of the urn U0 is taken to be a probability vector and can be thought to be the proportion of balls of each color/type we start with. Then P (A ball of color j is selected at the first trial | U0) = U0,j. We consider the replacement matrix R to be an infinite dimensional stochastic matrix. At each step n ≥ 1, the same procedure as that of Polya Urn Scheme is repeated. Let Un be the row vector denoting the “number” of balls of different colors at time n.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

2 / 17

SLIDE 7

Genaralization of the Polya Urn scheme to infinitely many colors We introduce an urn with infinite but countably many colors/types of balls indexed by Z. In this case, the so called “uniform” selection of balls does not make sense. The intial configuration of the urn U0 is taken to be a probability vector and can be thought to be the proportion of balls of each color/type we start with. Then P (A ball of color j is selected at the first trial | U0) = U0,j. We consider the replacement matrix R to be an infinite dimensional stochastic matrix. At each step n ≥ 1, the same procedure as that of Polya Urn Scheme is repeated. Let Un be the row vector denoting the “number” of balls of different colors at time n.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

2 / 17

SLIDE 8

Fundamental Recursion If the chosen ball turns out to be of jth color, then Un+1 is given by the equation Un+1 = Un + Rj where Rj is the jth row of the matrix R. This can also be written in the matrix notation as

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

3 / 17

SLIDE 9

Fundamental Recursion If the chosen ball turns out to be of jth color, then Un+1 is given by the equation Un+1 = Un + Rj where Rj is the jth row of the matrix R. This can also be written in the matrix notation as Un+1 = Un + In+1R (1)

where In = (. . . , In,−1, In,0, In,1 . . .) where In,i = 1 for i = j and 0 elsewhere. We study this process for the replacement matrices R which arise out of the Random Walks on Z.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

3 / 17

SLIDE 10

We can generalize this process to general graphs on Rd, d ≥ 1. Let G = (V, E) be a connected graph on Rd with vertex set V which is countably

infinite. The edges are taken to be bi-directional and there exists m ∈ N such

that d(v) = m for every v ∈ V. Let the distribution of X1 be given by P (X1 = v) = p(v) for v ∈ B where | B |< ∞. (2) where

v∈B

p(v) = 1. Let Sn =

n

i=1

Xi. Let R be the matrix/operator corresponding to the random walk Sn and the urn process evolve according to R. In this case, the configuration Un of the process is a row vector with co-ordinates indexed by V. The dynamics is similar to that in one-dimension, that is an element is drawn at random, its type noted and returned to the urn. If the vth type is selected at the n + 1 st trial, then Un+1 = Un + evR (3) where ev is a row vector with 1 at the vth co-ordinate and zero elsewhere.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

4 / 17

SLIDE 11

We note the following, for all d ≥ 1

v∈V

Un,v = n + 1.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

5 / 17

SLIDE 12

We note the following, for all d ≥ 1

v∈V

Un,v = n + 1. Hence

Un n+1 is a random probability vector. For every ω ∈ Ω, we can

define a random d-dimensional vector Tn(ω) with law Un(ω)

n+1 .

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

5 / 17

SLIDE 13

We note the following, for all d ≥ 1

v∈V

Un,v = n + 1. Hence

Un n+1 is a random probability vector. For every ω ∈ Ω, we can

define a random d-dimensional vector Tn(ω) with law Un(ω)

n+1 .

Also

(E[Un,v])v∈V n+1

is a probability vector. Therefore we can define a random vector Zn with law

(E[Un,v])v∈V n+1

.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

5 / 17

SLIDE 14

Previous work Literature is available only for finitely many types/ colors. It is known that the asymptotic behavior of the urn model depends on the qualitative properties (transience or recurrence) of the underlying Markov Chain of the replacement matrix. Svante Janson, Stochastic Processes, 2004.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

6 / 17

SLIDE 15

Previous work Literature is available only for finitely many types/ colors. It is known that the asymptotic behavior of the urn model depends on the qualitative properties (transience or recurrence) of the underlying Markov Chain of the replacement matrix. Svante Janson, Stochastic Processes, 2004. Svante Janson, Probab Theory and Related Fields, 2006.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

6 / 17

SLIDE 16

Previous work Literature is available only for finitely many types/ colors. It is known that the asymptotic behavior of the urn model depends on the qualitative properties (transience or recurrence) of the underlying Markov Chain of the replacement matrix. Svante Janson, Stochastic Processes, 2004. Svante Janson, Probab Theory and Related Fields, 2006. Arup Bose, Amites Dasgupta, Krishanu Maulik , Bernoulli, 2009.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

6 / 17

SLIDE 17

Previous work Literature is available only for finitely many types/ colors. It is known that the asymptotic behavior of the urn model depends on the qualitative properties (transience or recurrence) of the underlying Markov Chain of the replacement matrix. Svante Janson, Stochastic Processes, 2004. Svante Janson, Probab Theory and Related Fields, 2006. Arup Bose, Amites Dasgupta, Krishanu Maulik , Bernoulli, 2009. Arup Bose, Amites Dasgupta, Krishanu Maulik, Journal of Applied Probability, 2009.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

6 / 17

SLIDE 18

Previous work Literature is available only for finitely many types/ colors. It is known that the asymptotic behavior of the urn model depends on the qualitative properties (transience or recurrence) of the underlying Markov Chain of the replacement matrix. Svante Janson, Stochastic Processes, 2004. Svante Janson, Probab Theory and Related Fields, 2006. Arup Bose, Amites Dasgupta, Krishanu Maulik , Bernoulli, 2009. Arup Bose, Amites Dasgupta, Krishanu Maulik, Journal of Applied Probability, 2009. Amites Dasgupta, Krishanu Maulik, preprint.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

6 / 17

SLIDE 19

Previous work Literature is available only for finitely many types/ colors. It is known that the asymptotic behavior of the urn model depends on the qualitative properties (transience or recurrence) of the underlying Markov Chain of the replacement matrix. Svante Janson, Stochastic Processes, 2004. Svante Janson, Probab Theory and Related Fields, 2006. Arup Bose, Amites Dasgupta, Krishanu Maulik , Bernoulli, 2009. Arup Bose, Amites Dasgupta, Krishanu Maulik, Journal of Applied Probability, 2009. Amites Dasgupta, Krishanu Maulik, preprint.

T. W. Mullikan , Transactions of American Mathematical Society, 1963.
D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

6 / 17

SLIDE 20

Previous work Literature is available only for finitely many types/ colors. It is known that the asymptotic behavior of the urn model depends on the qualitative properties (transience or recurrence) of the underlying Markov Chain of the replacement matrix. Svante Janson, Stochastic Processes, 2004. Svante Janson, Probab Theory and Related Fields, 2006. Arup Bose, Amites Dasgupta, Krishanu Maulik , Bernoulli, 2009. Arup Bose, Amites Dasgupta, Krishanu Maulik, Journal of Applied Probability, 2009. Amites Dasgupta, Krishanu Maulik, preprint.

T. W. Mullikan , Transactions of American Mathematical Society, 1963.

Shu-Teh C. Moy , The Annals of Mathematical Statistics, 1966.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

6 / 17

SLIDE 21

Previous work Literature is available only for finitely many types/ colors. It is known that the asymptotic behavior of the urn model depends on the qualitative properties (transience or recurrence) of the underlying Markov Chain of the replacement matrix. Svante Janson, Stochastic Processes, 2004. Svante Janson, Probab Theory and Related Fields, 2006. Arup Bose, Amites Dasgupta, Krishanu Maulik , Bernoulli, 2009. Arup Bose, Amites Dasgupta, Krishanu Maulik, Journal of Applied Probability, 2009. Amites Dasgupta, Krishanu Maulik, preprint.

T. W. Mullikan , Transactions of American Mathematical Society, 1963.

Shu-Teh C. Moy , The Annals of Mathematical Statistics, 1966. Shu-Teh C. Moy , Journal of Mathematics and Mechanics, 1967.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

6 / 17

SLIDE 22

Main Result

Theorem

Let the process evolve according to a random walk on Rd with bounded

increments. Let the process begin with a single ball of type 0. For

X1 =

X(1)

1 , X(2) 1

. . . X(d)

1

, let µ =
E[X(1)

1 ], E[X(2) 1 ], . . . E[X(d) 1 ]

and

Σ = [σij]d×d where σi,j = E[X(i)

1 X(j) 1 ]. Let B be such that Σ is positive

definite. Then

Zn − µ log n √log n

d

− → N(0, Σ) as n → ∞ (4) where N(0, Σ) denotes the d-dimensional Gaussian with mean vector 0 and variance-covariance matrix Σ. Furthermore there exists a subsequence {nk} such that as k → ∞ almost surely Tnk − µ log n √log n

d

− → N(0, Σ) (5)

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

7 / 17

SLIDE 23

Interesting Examples

Corollary

Let d ≥ 1 and we consider the SSRW. Let the process begin with a single ball

f type 0. If Zn be the random d-dimensional vector corresponding to the

probability distribution

(E[Un,v])v∈Zd n+1

, then Zn √log n

d

− → N

0, d−1Id
as n → ∞

(6) where Id is the d-dimensional identity matrix. Furthemore, there exists a subsequence {nk} such that almost surely as k → ∞ Tnk √nk

d

− → N

0, d−1Id
.

(7)

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

8 / 17

SLIDE 24

Corollary

Let d = 1 and P (X1 = 1) = 1. Let U0 = 1{0}. If Zn be the random variable corresponding to the probability mass function (E[Un,k])k∈Z

n+1

, then Zn − log n √log n

d

− → N(0, 1) as n → ∞. (8) Also there exists a subsequence nk such that almost surely as k → ∞ Tnk − log nk √nk

d

− → N(0, 1). (9)

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

9 / 17

SLIDE 25

Figure: Triangular Lattice

Corollary

Let the urn model evolve according to the random walk on triangular lattice

n R2 and the process begin with a single particle of type 0, then as n → ∞

Zn √log n

d

− → N

0, 1

2I2

.

(10)

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

10 / 17

SLIDE 26

Corollary (continued)

Furthermore, there exists a subsequence {nk} such that as k → ∞, Tnk √log nk

d

− → N

0, 1

2I2

(11)
D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

11 / 17

SLIDE 27

Conclusion The SSRW is recurrent for d ≤ 2 and transient for d ≥ 3.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

12 / 17

SLIDE 28

Conclusion The SSRW is recurrent for d ≤ 2 and transient for d ≥ 3. In both cases, with a scaling of √log n the asymptotic behavior of the models are similar.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

12 / 17

SLIDE 29

Conclusion The SSRW is recurrent for d ≤ 2 and transient for d ≥ 3. In both cases, with a scaling of √log n the asymptotic behavior of the models are similar. On Z, the random walks are recurrent or transient depending on E[X1] = 0 or not. Asypmtotically both behave similarly upto centering and scaling.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

12 / 17

SLIDE 30

Conclusion The SSRW is recurrent for d ≤ 2 and transient for d ≥ 3. In both cases, with a scaling of √log n the asymptotic behavior of the models are similar. On Z, the random walks are recurrent or transient depending on E[X1] = 0 or not. Asypmtotically both behave similarly upto centering and scaling. We conjecture that in the infinite type/ color case, the asymptotic behavior of the process is not determined completely by the underlying Markov Chain of the operator, but by the qualitative properties of the underlying graph.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

12 / 17

SLIDE 31

Proof of the Main Theorem We present the proof for SSRW on d = 2 for notational simplicity. We use the martingale methods for the proof.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

13 / 17

SLIDE 32

Proof of the Main Theorem We present the proof for SSRW on d = 2 for notational simplicity. We use the martingale methods for the proof. For every t = (t1, t2) ∈ R2, e(t) = 1

4

u∈N(0)

eu,t is an eigen value for the operator R where 0 stands for the origin in Z2 and ., . stands for the inner product.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

13 / 17

SLIDE 33

Proof of the Main Theorem We present the proof for SSRW on d = 2 for notational simplicity. We use the martingale methods for the proof. For every t = (t1, t2) ∈ R2, e(t) = 1

4

u∈N(0)

eu,t is an eigen value for the operator R where 0 stands for the origin in Z2 and ., . stands for the inner product. The corresponding right eigen vectors are x (t) = (xt(v))v∈Z2 where xt(v) = et,v.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

13 / 17

SLIDE 34

Proof of the Main Theorem We present the proof for SSRW on d = 2 for notational simplicity. We use the martingale methods for the proof. For every t = (t1, t2) ∈ R2, e(t) = 1

4

u∈N(0)

eu,t is an eigen value for the operator R where 0 stands for the origin in Z2 and ., . stands for the inner product. The corresponding right eigen vectors are x (t) = (xt(v))v∈Z2 where xt(v) = et,v. We have noted earlier that

Un n+1 is a random probability vector.

The moment generating function for this vector is given by Un.x(t)

n+1

for every t ∈ R2.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

13 / 17

SLIDE 35

Proof of the Main Theorem We present the proof for SSRW on d = 2 for notational simplicity. We use the martingale methods for the proof. For every t = (t1, t2) ∈ R2, e(t) = 1

4

u∈N(0)

eu,t is an eigen value for the operator R where 0 stands for the origin in Z2 and ., . stands for the inner product. The corresponding right eigen vectors are x (t) = (xt(v))v∈Z2 where xt(v) = et,v. We have noted earlier that

Un n+1 is a random probability vector.

The moment generating function for this vector is given by Un.x(t)

n+1

for every t ∈ R2. Using (1), it can be shown that Mn(t) =

Un.x(t) Πn(e(t)) is a non-negative

martingale, where Πn(β) =

n

j=1

(1 + β j ).

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

13 / 17

SLIDE 36

Since we begin with one element of type 0, E

Mn(t)
= Πn (e(t)) .
D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

14 / 17

SLIDE 37

Since we begin with one element of type 0, E

Mn(t)
= Πn (e(t)) .

(12) Let us denote by En the expectation vector (E[Un,v])v∈Z2. The moment generating function for this vector is En.x(t)

n+1

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

14 / 17

SLIDE 38

Since we begin with one element of type 0, E

Mn(t)
= Πn (e(t)) .

(12) Let us denote by En the expectation vector (E[Un,v])v∈Z2. The moment generating function for this vector is En.x(t)

n+1

We will show that for a suitable δ > 0 and for all t ∈ [−δ, δ]2

En. x(

t √log n)

n + 1 − → e

t2 2 4

(13) where for all x ∈ R2, x 2 denontes the l2 norm.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

14 / 17

SLIDE 39

Since we begin with one element of type 0, E

Mn(t)
= Πn (e(t)) .

(12) Let us denote by En the expectation vector (E[Un,v])v∈Z2. The moment generating function for this vector is En.x(t)

n+1

We will show that for a suitable δ > 0 and for all t ∈ [−δ, δ]2

En. x(

t √log n)

n + 1 − → e

t2 2 4

(13) where for all x ∈ R2, x 2 denontes the l2 norm. We know that

En. x (tn) = Πn (e (tn))

(14) where tn =

t √log n.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

14 / 17

SLIDE 40

We use the following fact due to Euler, 1 Γ(β + 1) = lim

n→∞

Πn(β) nβ except for β non-negative integer.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

15 / 17

SLIDE 41

We use the following fact due to Euler, 1 Γ(β + 1) = lim

n→∞

Πn(β) nβ except for β non-negative integer. It is easy known that this convergence is uniform for all β ∈ [1 − η, 1 + η] for a suitable choice of η.

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

15 / 17

SLIDE 42

We use the following fact due to Euler, 1 Γ(β + 1) = lim

n→∞

Πn(β) nβ except for β non-negative integer. It is easy known that this convergence is uniform for all β ∈ [1 − η, 1 + η] for a suitable choice of η. Due to the uniform convergence, it follows immediately that ∀t ∈ [−δ, δ]2 lim

n→∞

Πn (e (tn)) ne(tn)/Γ(e (tn) + 1) = 1. (15)

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

15 / 17

SLIDE 43

Simplifying the left hand side of [13] we get Πn (e (tn)) n + 1 (16)

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

16 / 17

SLIDE 44

Simplifying the left hand side of [13] we get Πn (e (tn)) n + 1 (16) It is enough to show that lim

n→∞ − log(n + 1) + e (tn) log n − log(Γ(e (tn) + 1))

= t 2

2

4 . (17)

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

16 / 17

SLIDE 45

Simplifying the left hand side of [13] we get Πn (e (tn)) n + 1 (16) It is enough to show that lim

n→∞ − log(n + 1) + e (tn) log n − log(Γ(e (tn) + 1))

= t 2

2

4 . (17) Expanding e (tn) into power series and noting that Γ(x) is continuous as a function of x we can prove (17).

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

16 / 17

SLIDE 46

Thank You!

D. Thacker (Indian Statistical Institute)

On P´

lya Urn Schemes with Infinitelay Many Colors

17 / 17