the decoupling assumption in large stochastic … naughty example the dot is the starting point,...

37
Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion The decoupling assumption in large stochastic system analysis Talk at ECLT Andrea Marin 1 1 Dipartimento di Scienze Ambientali, Informatica e Statistica Università Ca’ Foscari Venezia, Italy (University of Venice, Italy) The decoupling assumption ECLT, 2016 1 / 29

Upload: dangkhuong

Post on 05-Apr-2018

216 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

The decoupling assumption in large stochasticsystem analysis

Talk at ECLT

Andrea Marin1

1Dipartimento di Scienze Ambientali, Informatica e StatisticaUniversità Ca’ Foscari Venezia, Italy

(University of Venice, Italy) The decoupling assumption ECLT, 2016 1 / 29

Page 2: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Outline

1 Motivation

2 Mean field and decoupling assumption

3 Product-forms and decoupling assumption

4 Conclusion

(University of Venice, Italy) The decoupling assumption ECLT, 2016 2 / 29

Page 3: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Section 1

Motivation

(University of Venice, Italy) The decoupling assumption ECLT, 2016 3 / 29

Page 4: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

What is the decoupling assumption?

It is normally underlying many analyses of stochastic systemsI Spreading of information in networksI Spreading of diseasesI Analysis of wireless and cabled communication networksI . . .

Why?I Without the decoupling assumption the system would be too

complicated to study

(University of Venice, Italy) The decoupling assumption ECLT, 2016 4 / 29

Page 5: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Running example

Example taken from the paper Discrete Markov chain approach tocontact-based disease spreading in complex networks by Gomezet al (2010), (130 citations)Goals of the paper:

I provide a model for contact-based disease spreadingI determine some values for the model parameters that characterise

the type of spreading (e.g., is it epidemic?)

(University of Venice, Italy) The decoupling assumption ECLT, 2016 5 / 29

Page 6: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

The model

N nodes connected according to a adjacency matrix R = (rij)where 0 ≤ rij ≤ 1 is the probability of node i to be in contact withnode jEach node can be in one of the following two states:

I susceptible (S)I infected (I)

The edge of the graph is a connection along which the infectionspreadsAt each time slot each infected node makes λ (independent) trialsto transmits the disease to its neighbour with probability β per timeunitµ is the rate at which a node moves from infected to susceptible

(University of Venice, Italy) The decoupling assumption ECLT, 2016 6 / 29

Page 7: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Example of dynamics

(University of Venice, Italy) The decoupling assumption ECLT, 2016 7 / 29

Page 8: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Example of dynamics

(University of Venice, Italy) The decoupling assumption ECLT, 2016 7 / 29

Page 9: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Example of dynamics

(University of Venice, Italy) The decoupling assumption ECLT, 2016 7 / 29

Page 10: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Example of dynamics

(University of Venice, Italy) The decoupling assumption ECLT, 2016 7 / 29

Page 11: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

What would we like to understand? Transient vs.Stationary analysis

Given an initial state, which is the probability of a certain(aggregated) state after n steps?

I Transient analysis, finite time horizonGiven an initial state, which is the probability of a certain(aggregated) state when n→∞?

I Stationary analysisI Does the system reach an equilibrium?I Does it depend on the initial state?

In the running example the authors focus on the stationaryanalysis

I Problem: in a network of N nodes we obtain a Markov chain of 2N

states whose stationary analysis has the computational cost ofO(23N)

(University of Venice, Italy) The decoupling assumption ECLT, 2016 8 / 29

Page 12: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

The model

pi(t) probability of node i of being infectedβ is the intensity of infection spreadingµ recovery rateqi(t) probability of node i not being infected by any of itsneighbours

pi(t + 1) = (1− qi(t))(1− pi(t)) + (1− µ)pi(t) + µ(1− qi(t))pi(t)

qi(t) =

N∏j=1

(1− βrijpj(t))

(University of Venice, Italy) The decoupling assumption ECLT, 2016 9 / 29

Page 13: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

The model

pi(t) probability of node i of being infectedβ is the intensity of infection spreadingµ recovery rateqi(t) probability of node i not being infected by any of itsneighbours

pi(t + 1) = (1− qi(t))(1− pi(t)) + (1− µ)pi(t) + µ(1− qi(t))pi(t)

qi(t) =

N∏j=1

(1− βrijpj(t))

(University of Venice, Italy) The decoupling assumption ECLT, 2016 9 / 29

Page 14: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

The model

pi(t) probability of node i of being infectedβ is the intensity of infection spreadingµ recovery rateqi(t) probability of node i not being infected by any of itsneighbours

pi(t + 1) = (1− qi(t))(1− pi(t)) + (1− µ)pi(t) + µ(1− qi(t))pi(t)

qi(t) =

N∏j=1

(1− βrijpj(t))

(University of Venice, Italy) The decoupling assumption ECLT, 2016 9 / 29

Page 15: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

The model

pi(t) probability of node i of being infectedβ is the intensity of infection spreadingµ recovery rateqi(t) probability of node i not being infected by any of itsneighbours

pi(t + 1) = (1− qi(t))(1− pi(t)) + (1− µ)pi(t) + µ(1− qi(t))pi(t)

qi(t) =

N∏j=1

(1− βrijpj(t))

(University of Venice, Italy) The decoupling assumption ECLT, 2016 9 / 29

Page 16: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

The model

pi(t) probability of node i of being infectedβ is the intensity of infection spreadingµ recovery rateqi(t) probability of node i not being infected by any of itsneighbours

pi(t + 1) = (1− qi(t))(1− pi(t)) + (1− µ)pi(t) + µ(1− qi(t))pi(t)

qi(t) =

N∏j=1

(1− βrijpj(t))

(University of Venice, Italy) The decoupling assumption ECLT, 2016 9 / 29

Page 17: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

The model

pi(t) probability of node i of being infectedβ is the intensity of infection spreadingµ recovery rateqi(t) probability of node i not being infected by any of itsneighbours

pi(t + 1) = (1− qi(t))(1− pi(t)) + (1− µ)pi(t) + µ(1− qi(t))pi(t)

qi(t) =

N∏j=1

(1− βrijpj(t))

Where is the decoupling assumption?

(University of Venice, Italy) The decoupling assumption ECLT, 2016 9 / 29

Page 18: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

From the paper. . .

The formulation so far relies on the assumption that theprobabilities of being infected pi are independent randomvariables. This hypothesis turns out to be valid in the vastmajority of complex networks because the inherenttopological disorder makes dynamical correlations notpersistent.

Is that enough?Afterward the pi are computed as functions of β and µ by solving afixed point iteration scheme

(University of Venice, Italy) The decoupling assumption ECLT, 2016 10 / 29

Page 19: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Section 2

Mean field and decoupling assumption

(University of Venice, Italy) The decoupling assumption ECLT, 2016 11 / 29

Page 20: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Roadmap

We consider two analysis approaches:I Mean field modelsI Product-form models

We study the decoupling assumption for the two settings:I TransientI Stationary

Transient StationaryMean field ?? ??

Product-forms ?? ??

(University of Venice, Italy) The decoupling assumption ECLT, 2016 12 / 29

Page 21: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Mean field in a nutshell

Example: infection spreading1

N individuals who can be in one of the following three states:I D→ Dormant (infected but with no visible symptoms)I A→ Active (infected with visible symptoms)I S→ Susceptible

Discrete time setting

1Taken from A class of mean field interaction models for computer andcommunication systems, by Le Boudec et al.

(University of Venice, Italy) The decoupling assumption ECLT, 2016 13 / 29

Page 22: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Transition rules (from dormant)

D→ proportion dormant elements, A→ proportion of active elements,S→ proportion of susceptible elements

Recovering with probability δD

Activation with probability NDN−1N λ

(University of Venice, Italy) The decoupling assumption ECLT, 2016 14 / 29

Page 23: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Transition rules (from active)

Recovering with probability δA

Activation with probability DN

h+DN β

(University of Venice, Italy) The decoupling assumption ECLT, 2016 15 / 29

Page 24: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Transition rules (from susceptible)

Exogenous infection α0

Infection with probability rDN

(University of Venice, Italy) The decoupling assumption ECLT, 2016 16 / 29

Page 25: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Transition rules (from susceptible)

Exogenous activation α

Mean field: under some (mild) conditions, for N →∞ theprobabilistic model’s behaviour coincides almost surely with thetrajectory of the solution of ODE system for any finite time horizonProbabilistic→ deterministic

(University of Venice, Italy) The decoupling assumption ECLT, 2016 17 / 29

Page 26: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

The ODE associated with the system

The drift for each variable D, A, S is given by:∆D = −DδD − 2D2λ− Aβ D(t)

h+D + S(α0 + rD)

∆A = 2D2λ+ Aβ Dh+D − AδA + Sα

∆S = DδD + AδA − S(α0 + rD)− Sα

(University of Venice, Italy) The decoupling assumption ECLT, 2016 18 / 29

Page 27: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Good news!

The decoupling assumption holds in transient regime!Consequences:

I We can focus on a single individualI Its behaviour is probabilistic but it interacts with a deterministic

environment as N →∞I The analysis corresponds to the transient analysis of a

time-inhomogeneous Markov chainI Idea: at each simulation step the environment changes the

transition probabilities of the Markov chain associated with a singleindividual (only three states!)

Transient StationaryMean field OK ??

Product-forms ?? ??

(University of Venice, Italy) The decoupling assumption ECLT, 2016 19 / 29

Page 28: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

What about the stationary behaviour?

We can set the drift to 0 and look for a solution for D, S, AIn general the ODE may admit more fixed points

I In this case the decoupling assumption in stationary regime doesnot hold

Is proving the uniqueness of the fixed point enough?

(University of Venice, Italy) The decoupling assumption ECLT, 2016 20 / 29

Page 29: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

A nice example

The dot is the starting point, while the cross is the fixed point solution.Taken from A class of mean field interaction models for computer and communication systems, by le Boudec et al.

(University of Venice, Italy) The decoupling assumption ECLT, 2016 21 / 29

Page 30: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

A naughty example

The dot is the starting point, while the cross is the fixed point solution.Taken from A class of mean field interaction models for computer and communication systems, by le Boudec et al.

(University of Venice, Italy) The decoupling assumption ECLT, 2016 22 / 29

Page 31: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Not so good news

The decoupling assumption is valid in stationary regime ifI We have a unique fixed point of the ODEI All the trajectories of the system converge to the fixed pointI These properties depend on the specific ODE, hard to find general

results

This is usually extremely hard to prove

Transient StationaryMean field OK Sometimes, hard to prove

Product-forms ?? ??

(University of Venice, Italy) The decoupling assumption ECLT, 2016 23 / 29

Page 32: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Section 3

Product-forms and decoupling assumption

(University of Venice, Italy) The decoupling assumption ECLT, 2016 24 / 29

Page 33: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

When is a stochastic model in product-form?

Setting: continuous timeN interacting individuals

I Sn the state space of nI Joint state space: S ⊆ S1 × S2 × · · · × SN

π(s) be the stationary probability of state s ∈ S:

π(s) ∝N∏

n=1

gn(sn)

gn(sn) is interpreted as the stationary probability of individual nisolated and re parameterised to take into account the interactionswith the other individualsProduct-form 6= stochastic independence since for t ∈ R, ingeneral:

π(s, t) 6=N∏

n=1

gn(sn, t)

(University of Venice, Italy) The decoupling assumption ECLT, 2016 25 / 29

Page 34: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Example: migration process

N individuals clustered in J coloniesEach colony has ni individuals, with

∑Ni=1 ni = N

State of the system n = (n1, . . . , nJ)

Tjk is the operator that moves one individual from colony j, withnj > 0 to colony nk

Colony connections are modelled by a graph with adjacencymatrix R = (rij), rij ∈ {0, 1}The migration process is regulated by the law:

q(n,Tjkn) = rijλjkΦj(nj)

with Φj(0) = 0.

(University of Venice, Italy) The decoupling assumption ECLT, 2016 26 / 29

Page 35: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Product-form

The stationary probability of observing state n is:

π(n) ∝J∏

j=1

αnjj∏nj

r=1 Φj(r)

αj is a non-trivial solution of αj∑

k λjkrjk =∑

k αkλkjrkj

We only need the graph to be irreducibleI No limiting assumptions on the structure or on the population

Transient StationaryMean field OK Sometimes, hard to prove

Product-forms NO OK

(University of Venice, Italy) The decoupling assumption ECLT, 2016 27 / 29

Page 36: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Section 4

Conclusion

(University of Venice, Italy) The decoupling assumption ECLT, 2016 28 / 29

Page 37: The decoupling assumption in large stochastic … naughty example The dot is the starting point, while the cross is the fixed point solution. Taken from A class of mean field interaction

Motivation Mean field and decoupling assumption Product-forms and decoupling assumption Conclusion

Conclusion

The decoupling assumption is often needed to make the modelstractableHandling the decoupling assumption correctly is not trivialMean field vs. Product-forms:

I Mean field: less restrictions on the model, easy to handle thetransient, unclear when the limiting approximation is good.Stationary analyses must be handled carefully;

I Product-forms: models must fulfil some conditions, useful in thestationary regime, decoupling assumption holds, almost no resultsin the transient regime

(University of Venice, Italy) The decoupling assumption ECLT, 2016 29 / 29