GraphDiseaseSpreadModels-Threshold&FirefighterCapeTown6-10-07

Graph-theoretical Problems Arising from Defending Against Bioterrorism and Controlling the Spread of Fires Fred Roberts, DIMACS

Mathematical Models of Disease Spread Mathematical models of infectious diseases go back to Daniel Bernoulli’s mathematical analysis of smallpox in 1760.

Understanding infectious systems requires being able to reason about highly complex biological systems, with hundreds of demographic and epidemiological variables . Intuition alone is insufficient to fully understand the dynamics of such systems. smallpox

Experimentation or field trials are often prohibitively expensive or unethical and do not always lead to fundamental understanding. Therefore, mathematical modeling becomes an important experimental and analytical tool.

Mathematical models have become important tools in analyzing the spread and control of infectious diseases, especially when combined with powerful, modern computer methods for analyzing and/or simulating the models.

Great concern about the deliberate introduction of diseases by bioterrorists has led to new challenges for mathematical modelers. anthrax

Great concern about possibly devastating new diseases like avian influenza has also led to new challenges for mathematical modelers.

Hundreds of math. models since Bernoulli’s models of smallpox have: highlighted concepts like core population in STD’s;

Made explicit concepts such as herd immunity for vaccination policies;

Led to insights about drug resistance, rate of spread of infection, epidemic trends, effects of different kinds of treatments.

The size and overwhelming complexity of modern epidemiological problems -- and in particular the defense against bioterrorism -- calls for new approaches and tools.

The Methods of Mathematical Epidemiology Statistical Methods long history in epidemiology changing due to large data sets involved Dynamical Systems model host-pathogen systems, disease spread difference and differential equations little systematic use of today’s powerful computational methods

The Methods of Mathematical Epidemiology Probabilistic Methods stochastic processes, random walks, percolation, Markov chain Monte Carlo methods simulation need to bring in more powerful computational tools

Discrete Mathematics Many fields of science, in particular molecular biology, have made extensive use of DM broadly defined.

Discrete Mathematics These tools have remained largely unused and unknown in epidemiology and even mathematical epidemiology. However, especially with the beginning of social network analysis, DM is becoming to be more widely used. Goals today: Introduce examples of discrete models Make some comments about the role of mathematical modeling

Models of the Spread and Control of Disease through Social Networks Diseases are spread through social networks. “ Contact tracing ” is an important part of any strategy to combat outbreaks of infectious diseases, whether naturally occurring or resulting from bioterrorist attacks. AIDS

The Model: Moving From State to State Social Network = Graph Vertices = People Edges = contact Let s i (t) give the state of vertex i at time t. Simplified Model: Two states: = susceptible, = infected (SI Model) Times are discrete: t = 0, 1, 2, … t=0

The Model: Moving From State to State More complex models: SI, SEI, SEIR, etc. S = susceptible, E = exposed, I = infected, R = recovered (or removed) measles SARS

Threshold Processes Irreversible k-Threshold Process : You change your state from to at time t+1 if at least k of your neighbors have state at time t. You never leave state . Disease interpretation? Infected if sufficiently many of your neighbors are infected. Special Case k = 1: Infected if any of your neighbors is infected.

Irreversible 2-Threshold Process t=0

Irreversible 2-Threshold Process t=1 t=0

Irreversible 2-Threshold Process t=1 t=2

Irreversible 3-Threshold Process t = 0 a e d c b f g

Irreversible 3-Threshold Process t = 0 a e d c b f g t = 1 a e d c b f g

Irreversible 3-Threshold Process t = 1 a e d c b f g a e d c b f g t = 2

Complications to Add to Model k = 1, but you only get infected with a certain probability. You are automatically cured after you are in the infected state for d time periods. A public health authority has the ability to “vaccinate” a certain number of vertices, making them immune from infection. Waiting for smallpox vaccination, NYC, 1947

Vaccination Strategies Mathematical models are very helpful in comparing alternative vaccination strategies. The problem is especially interesting if we think of protecting against deliberate infection by a bioterrorist.

Vaccination Strategies If you didn’t know whom a bioterrorist might infect, what people would you vaccinate to be sure that a disease doesn’t spread very much? (Vaccinated vertices stay at state regardless of the state of their neighbors.) Try odd cycles. Consider an irreversible 2-threshold process. Suppose your adversary has enough supply to infect two individuals. 5-cycle C 5

Vaccination Strategies One strategy: “Mass vaccination ” : Make everyone and immune in initial state. In 5-cycle C 5 , mass vaccination means vaccinate 5 vertices. This obviously works. In practice, vaccination is only effective with a certain probability, so results could be different. Can we do better than mass vaccination? What does better mean? If vaccine has no cost and is unlimited and has no side effects, of course we use mass vaccination.

Vaccination Strategies What if vaccine is in limited supply? Suppose we only have enough vaccine to vaccinate 2 vertices. two different vaccination strategies: Vaccination Strategy I Vaccination Strategy II V V V V

Vaccination Strategy I: Worst Case (Adversary Infects Two) Two Strategies for Adversary Adversary Strategy Ia Adversary Strategy Ib I I I I This assumes adversary doesn’t attack a vaccinated vertex. Problem is interesting if this could happen – or you encourage it to happen. V V V V

The “alternation” between your choice of a defensive strategy and your adversary’s choice of an offensive strategy suggests we consider the problem from the point of view of game theory. The Food and Drug Administration is studying the use of game-theoretic models in the defense against bioterrorism.

Vaccination Strategy I Adversary Strategy Ia I I t = 0 V V

Vaccination Strategy I Adversary Strategy Ia I I I I t = 0 t = 1 V V V V

Vaccination Strategy I Adversary Strategy Ia I I I I t = 1 t = 2 V V V V

Vaccination Strategy I Adversary Strategy Ib I I t = 0 V V

Vaccination Strategy I Adversary Strategy Ib I I I I t = 0 t = 1 V V V V

Vaccination Strategy I Adversary Strategy Ib I I I I t = 1 t = 2 V V V V

Vaccination Strategy II: Worst Case (Adversary Infects Two) Two Strategies for Adversary Adversary Strategy IIa Adversary Strategy IIb I I I I V V V V

Vaccination Strategy II Adversary Strategy IIa I I t = 0 V V

Vaccination Strategy II Adversary Strategy IIa I I t = 0 I I t = 1 V V V V

Vaccination Strategy II Adversary Strategy IIa I I t = 1 I I t = 2 V V V V

Vaccination Strategy II Adversary Strategy IIb I t = 0 I V V

Vaccination Strategy II Adversary Strategy IIb I t = 0 I t = 1 V V V V

Vaccination Strategy II Adversary Strategy IIb I t = 1 I t = 2 V V V V

Conclusions about Strategies I and II Vaccination Strategy II never leads to more than two infected individuals, while Vaccination Strategy I sometimes leads to three infected individuals (depending upon strategy used by adversary). Thus, Vaccination Strategy II is better. More on vaccination strategies later.

The Saturation Problem Attacker’s Problem : Given a graph, what subsets S of the vertices should we plant a disease with so that ultimately the maximum number of people will get it? Economic interpretation: What set of people do we place a new product with to guarantee “saturation” of the product in the population? Defender’s Problem : Given a graph, what subsets S of the vertices should we vaccinate to guarantee that as few people as possible will be infected?

k-Conversion Sets Attacker’s Problem: Can we guarantee that ultimately everyone is infected? Irreversible k-Conversion Set : Subset S of the vertices that can force an irreversible k-threshold process to the situation where every state s i (t) = ? Comment: If we can change back from to at least after awhile, we can also consider the Defender’s Problem: Can we guarantee that ultimately no one is infected, i.e., all s i (t) = ?

What is an irreversible 2-conversion set for the following graph? x 1 x 2 x 3 x 4 x 6 x 5

x 1 , x 3 is an irreversible 2-conversion set. t = 0 x 1 x 2 x 3 x 4 x 6 x 5

Irreversible k-Conversion Sets in Regular Graphs G is r-regular if every vertex has degree r. Degree = number of neighbors. Set of vertices is independent if there are no edges. C 5 is 2-regular. The two circled vertices form an independent set. No set of three vertices is independent. The largest independent set has size floor[5/2] = 2.

Irreversible k-Conversion Sets in Regular Graphs G is r-regular if every vertex has degree r. Set of vertices is independent if there are no edges. Theorem (Dreyer 2000): Let G = (V,E) be a connected r-regular graph and D be a set of vertices. Then D is an irreversible r-conversion set iff V-D is an independent set. Note: same r

k-Conversion Sets in Regular Graphs Corollary (Dreyer 2000): The size of the smallest irreversible 2- conversion set in C n is ceiling[n/2].

k-Conversion Sets in Regular Graphs Corollary (Dreyer 2000): The size of the smallest irreversible 2- conversion set in C n is ceiling[n/2]. C 5 is 2-regular. The smallest irreversible 2-conversion set has three vertices: the red ones.

k-Conversion Sets in Regular Graphs Corollary (Dreyer 2000): The size of the smallest irreversible 2- conversion set in C n is ceiling[n/2]. Proof : C n is 2-regular. The largest independent set has size floor[n/2]. Thus, the smallest D so that V-D is independent has size ceiling[n/2].

k-Conversion Sets in Regular Graphs Another Example: a e d c b f

k-Conversion Sets in Regular Graphs Another Example: This is 3- regular. Let k = 3. The largest independent set has 2 vertices. a e d c b f

k-Conversion Sets in Regular Graphs The largest independent set has 2 vertices. Thus, the smallest irreversible 3-conversion set has 6-2 = 4 vertices. The 4 red vertices form such a set. Each other vertex has three red neighbors. a e d c b f

Irreversible k-Conversion Sets in Graphs of Maximum Degree r Theorem (Dreyer 2000): Let G = (V,E) be a connected graph with maximum degree r and S be the set of all vertices of degree < r. If D is a set of vertices, then D is an irreversible r-conversion set iff S  D and V-D is an independent set.

How Hard is it to Find out if There is an Irreversible k-Conversion Set of Size at Most p? Problem IRREVERSIBLE k-CONVERSION SET : Given a positive integer p and a graph G, does G have an irreversible k-conversion set of size at most p? How hard is this problem?

Difficulty of Finding Irreversible Conversion Sets Problem IRREVERSIBLE k-CONVERSION SET : Given a positive integer p and a graph G, does G have an irreversible k-conversion set of size at most p? Theorem (Dreyer 2000): IRREVERSIBLE k-CONVERSION SET is NP-complete for fixed k > 2. (Whether or not it is NP-complete for k = 2 remains open.) Thus in technical CS terms, the problem is HARD.

Irreversible k-Conversion Sets in Trees

Irreversible k-Conversion Sets in Trees Tree = graph with: (1) no cycles (2) you can get from every vertex to every other vertex (connectedness)

Irreversible k-Conversion Sets in Trees The simplest case is when every internal vertex of the tree has degree > k. Leaf = vertex of degree 1; internal vertex = not a leaf. What is an irreversible 2-conversion set here?

Irreversible k-Conversion Sets in Trees The simplest case is when every internal vertex of the tree has degree > k. Leaf = vertex of degree 1; internal vertex = not a leaf. What is an irreversible 2-conversion set here? Do you know any vertices that have to be in such a set?

All leaves have to be in it. This will suffice. t = 0

All leaves have to be in it. This will suffice. t = 0 t = 1

All leaves have to be in it. This will suffice. t = 1 t = 2

Irreversible k-Conversion Sets in Trees So k = 2 is easy. What about k > 2? Also easy. Proposition (Dreyer 2000): Let T be a tree and every internal vertex have degree > k, where k > 1. Then the smallest irreversible k-conversion set has size equal to the number of leaves of the tree.

Irreversible k-Conversion Sets in Trees What if not every internal vertex has degree > k? If there is an internal vertex of degree < k, it will have to be in any irreversible k-conversion set and will never change sign. So, to every neighbor, this vertex v acts like a leaf, and we can break T into deg(v) subtrees with v a leaf in each. If every internal vertex has degree  k, one can obtain analogous results to those for the > k case by looking at maximal connected subsets of vertices of degree k.

Irreversible k-Conversion Sets in Trees What if not every internal vertex has degree > k? Question: Can you find an example where the set of leaves is not an irreversible k-conversion set?

Irreversible k-Conversion Sets in Trees What if not every internal vertex has degree > k? Question: Can you find an example where the set of leaves is not an irreversible k-conversion set? Yes: if a vertex has degree < k, even if it is not a leaf, it must be in every irreversible k-conversion set.

Irreversible k-Conversion Sets in Trees Dreyer presents an O(n) algorithm for finding the size of the smallest irreversible k-conversion set in a tree of n vertices. O(n) is considered very efficient.

Irreversible k-Conversion Sets in Special Graphs Studied for many special graphs. Let G(m,n) be the rectangular grid graph with m rows and n columns. G(3,4)

Toroidal Grids The toroidal grid T(m,n) is obtained from the rectangular grid G(m,n) by adding edges from the first vertex in each row to the last and from the first vertex in each column to the last. Toroidal grids are easier to deal with than rectangular grids because they form regular graphs: Every vertex has degree 4. Thus, we can make use of the results about regular graphs.

Irreversible4-Conversion Sets in Toroidal Grids Theorem (Dreyer 2000): In a toroidal grid T(m,n), the size of the smallest irreversible 4-conversion set is max{n(ceiling[m/2]), m(ceiling[n/2])} m or n odd mn/2 m, n even {

Part of the Proof : Recall that D is an irreversible 4-conversion set in a 4-regular graph iff V-D is independent. V-D independent means that every edge {u,v} in G has u or v in D. In particular, the ith row must contain at least ceiling[n/2] vertices in D and the ith column at least ceiling[m/2] vertices in D (alternating starting with the end vertex of the row or column). We must cover all rows and all columns, and so need at least max{n(ceiling[m/2]), m(ceiling[n/2])} vertices in an irreversible 4-conversion set.

Irreversible k-Conversion Sets for Rectangular Grids Let C k (G) be the size of the smallest irreversible k-conversion set in graph G. Theorem (Dreyer 2000): C 4 [G(m,n)] = 2m + 2n - 4 + floor[(m-2)(n-2)/2] Theorem (Flocchini, Lodi, Luccio, Pagli, and Santoro): C 2 [G(m,n)] = ceiling([m+n]/2)

Irreversible 3-Conversion Sets for Rectangular Grids For 3-conversion sets, the best we have are bounds: Theorem (Flocchini, Lodi, Luccio, Pagli, and Santoro): [(m-1)(n-1)+1]/3  C 3 [G(m,n)]  [(m-1)(n-1)+1]/3 +[3m+2n-3]/4 + 5 Finding the exact value is an open problem.

Irreversible Conversion Sets for Rectangular Grids Exact values are known for the size of the smallest irreversible k-conversion set for some special classes of graphs and some values of k: 2xn grids, 3xn grids, trees, etc.

Bounds on the Size of the Smallest Conversion Sets In general, it is difficult to get exact values for the size of the smallest irreversible k-conversion set in a graph. So, what about bounds? Sample result: Theorem (Dreyer, 2000): If G is an r-regular graph with n vertices, then C k (G)  (1 – r/2k)n for k  r  2k.

Vaccination Strategies Stephen Hartke worked on a different problem: Defender: can vaccinate v people per time period . Attacker: can only infect people at the beginning. Irreversible k-threshold model. What vaccination strategy minimizes number of people infected? Sometimes called the firefighter problem : alternate fire spread and firefighter placement. Usual assumption: k = 1. (We will assume this.) Variation: The vaccinator and infector alternate turns, having v vaccinations per period and i doses of pathogen per period. What is a good strategy for the vaccinator? Chapter in Hartke’s Ph.D. thesis at Rutgers (2004)

A Survey of Some Results on the Firefighter Problem Thanks to Kah Loon Ng DIMACS For the following slides, slightly modified by me

Mathematicians can be Lazy Different application. Different terminology Same mathematical model. measles

A Simple Model (k = 1) (v = 3)

Some questions that can be asked (but not necessarily answered!) Can the fire be contained? How many time steps are required before fire is contained? How many firefighters per time step are necessary? What fraction of all vertices will be saved (burnt)? Does where the fire breaks out matter? Fire starting at more than 1 vertex? Consider different graphs. Construction of (connected) graphs to minimize damage. Complexity/Algorithmic issues

Containing Fires in Infinite Grids L d Fire starts at only one vertex: d= 1: Trivial. d = 2: Impossible to contain the fire with 1 firefighter per time step

Containing Fires in Infinite Grids L d d = 2: Two firefighters per time step needed to contain the fire. 8 time steps 18 burnt vertices

Containing Fires in Infinite Grids L d …… d  3: Wang and Moeller (2002): If G is an r -regular graph, r – 1 firefighters per time step is always sufficient to contain any fire outbreak (at a single vertex) in G . ( r-regular : every vertex has r neighbors.) .….

Containing Fires in Infinite Grids L d d  3: In L d , every vertex has degree 2 d . Thus: 2 d -1 firefighters per time step are sufficient to contain any outbreak starting at a single vertex. Theorem (Hartke 2004): If d  3, 2 d – 2 firefighters per time step are not enough to contain an outbreak in L d. Thus, 2 d – 1 firefighters per time step is the minimum number required to contain an outbreak in L d and containment can be attained in 2 time steps.

Containing Fires in Infinite Grids L d Fire can start at more than one vertex. d = 2: Fogarty (2003): Two firefighters per time step are sufficient to contain any outbreak at a finite number of vertices. d  3: Hartke (2004): For any d  3 and any positive integer f, f firefighters per time step is not sufficient to contain all finite outbreaks in L d . In other words, for d  3 and any positive integer f , there is an outbreak such that f firefighters per time step cannot contain the outbreak.

Containing Fires in Infinite Grids L d The case of a different number of firefighters per time step. Let f(t) = number firefighters available at time t. Assume f(t) is periodic with period p f . Possible motivations for periodicity: Firefighters arrive in batches. Firefighters need to stay at a vertex for several time periods before redeployment.

Containing Fires in Infinite Grids L d The case of a different number of firefighters per time step. N f = f(1) + f(2) + … + f(p f ) R f = N f /p f (average number firefighters available per time period) Theorem (Ng and Raff 2006): If d =2 and f is periodic with period p f  1 and R f > 1.5, then an outbreak at any number of vertices can be contained at a finite number of vertices.

Saving Vertices in Finite Grids G Assumptions: 1 firefighter is deployed per time step Fire starts at one vertex Let MVS ( G , v ) = maximum number of vertices that can be saved in G if fire starts at v .

Saving Vertices in Finite Grids G

Algorithmic and Complexity Matters FIREFIGHTER: Instance: A rooted graph (G,u) and an integer p  1. Question: Is MVS(G,u)  p? That is, is there a finite sequence d 1 , d 2 , …, d t of vertices of G such that if the fire breaks out at u, then, 1. vertex d i is neither burning nor defended at time i 2. at time t, no undefended vertex is next to a burning vertex 3. at least p vertices are saved at the end of time t.

Algorithmic and Complexity Matters Theorem (MacGillivray and Wang, 2003): FIREFIGHTER is NP-complete. Thus, it is HARD in the sense of computer science.

Algorithmic and Complexity Matters Firefighting on Trees:

Algorithmic and Complexity Matters Greedy algorithm : For each v in V(T), define weight (v) = number descendants of v + 1 Algorithm: At each time step, place firefighter at vertex that has not been saved such that weight ( v ) is maximized.

Algorithmic and Complexity Matters Firefighting on Trees: 7 8 9 12 11 3 2 4 1 6 1 5 1 2 6 1 2 1 1 3 1 1 1 1 3 1 26 22

Algorithmic and Complexity Matters Greedy Optimal = 7 = 9

Algorithmic and Complexity Matters Theorem (Hartnell and Li, 2000): For any tree with one fire starting at the root and one firefighter to be deployed per time step, the greedy algorithm always saves more than ½ of the vertices that any algorithm saves.

More Realistic Models Many oversimplifications in both of our models. For instance: What if you stay infected (burning) only a certain number of days? What if you are not necessarily infective for the first few days you are sick? What if your threshold k for changes from to (uninfected to infected) changes depending upon how long you have been uninfected? measles

More Realistic Models Consider an irreversible process in which you stay in the infected state (state ) for d time periods after entering it and then go back to the uninfected state (state ). Consider an irreversible k-threshold process in which we vaccinate a person in state once k-1 neighbors are infected (in state ). Etc. – experiment with a variety of assumptions

More Realistic Models Our models are deterministic . How do probabilities enter? What if you only get infected with a certain probability if you meet an infected person? What if vaccines only work with a certain probability? What if the amount of time you remain infective exhibits a probability distribution?

Would Graph Theory help with an outbreak of Bovine TB?

What about an outbreak of Avian Flu?

Similar approaches using mathematical models have proven useful in public health and many other fields, to: make policy plan operations analyze risk compare interventions identify the cause of observed events So, why not for the spread of infectious disease?

Other Questions Can you use graph-theoretical models to analyze the effect of different quarantine strategies? Don’t forget diseases of plants.

There is much more analysis of a similar nature that can be done with mathematical models. Let your imagination run free!

GraphDiseaseSpreadModels-Threshold&FirefighterCapeTown6-10-07

More Related Content

Similar to GraphDiseaseSpreadModels-Threshold&FirefighterCapeTown6-10-07 (20)

More from webuploader (20)

Recently uploaded (20)

GraphDiseaseSpreadModels-Threshold&FirefighterCapeTown6-10-07