EE228a - Lecture 20 - Spring 2006 Concave Games, Learning in Games, Cooperative Games
|
|
- Rosalind Murphy
- 5 years ago
- Views:
Transcription
1 EE228a - Lecture 20 - Spring 2006 Concave Games, Learning in Games, Cooperative Games Jean Walrand - scribed by Michael Krishnan (mnk@berkeley.edu) Abstract In this lecture we extend our basic game theory concepts to more complex situations. We start with games in a continuous action space, looking at a specific class (concave games) that still have a Nash Equilibrium. We also go into repeated games and the effects of learning. Finally we look at cooperative games, where we must find a fair way to distribute a resource among cooperative players. After this we take a look at some interesting fixed-point theorems. 1 Concave Games 1.1 Motivation In many applications, the possible actions belong to a continuous set. For instance one chooses prices, transmission rates, or power levels. In such situations, one specifies reward functions instead of a matrix or rewards. 1.2 Preliminaries Figure 1: Best strategy curves for a game with a continuous action space: In the first, the discontinuity results in no Nash Equilibrium. In the second, the curvature of the curves results in multiple Nash Equilibria. In the third there is a unique Nash Equilibrium.[1] From figure 1, we can see that it is possible to have no Nash Equilibrium, several Nash Equilibria, or one unique Nash Equilibrium. We are interested in a subset of continuous games that have a unique Nash Equilibrium. It turns out that a set of games that can satisfy this criteria is concave games. 1
2 Definition 1 (Concave Game) We shall define a concave game as a game in which player i chooses an action x i R m i so that x = (x 1,..., x n) C, where C is a closed bounded convex set. The payoff function of player i is φ i(x), continuous in x and concave in x i for x C. Definition 2 (Nash Equilibrium of a Concave Game) We can then define the Nash Equilibrium to be x 0 C such that φ i(x 0 ) φ i(x 0 i, y i) y i such that (x 0 i, y i) C. (Here (x 0 i, y i) is the notation used for a vector in which the action of the i t h player is replaced by y i.) We will now show that concave games have a unique Nash Equilibrium. We begin by proving existence: Theorem 1 (Existence) A concave game always has at least one Nash Equilibrium Proof: Define ρ(x, y) := P n i=1 φi(x i, yi) for (x, y) C2. Define ρ (x) := max z Cρ(x, z). Define Γ(x) := {y C ρ(x, y) = ρ (x)} Intuitively, Γ(x) i is the strategy that player i would play given that the other players are playing x. Thus a fixed point x Γ(x) of Γ must be a Nash Equilibrium. If we know that Γ is convex and graph-continuous, then by Kakutani (proof in section 4.2), it has a fixed point. In general, it may be hard to show that Γ is convex and graph-continuous, so we will look at a specialized case. We will require that C = {x h(x) 0} where h( ) : R m R k is a vector of C 1, concave functions. We assume that C is bounded and that there is some x 0 C such that h(x 0 ) > 0 (strictly greater to avoid the trivial case). Also, φ j(x) has a continuous first derivative jφ j(x) w.r.t. x j. Thus, the objective functions are concave and the constraints are concave ( ) 0. The proof for uniqueness is harder. To simplify things, we will use a new definition: Definition 3 (Diagonally Strictly Concave) The function σ(x, r) := P n i=1 riφi(x), r Rn + is diagonally strictly concave(dsc) if (x 1 x 0 ) g(x 0, r)+(x 0 x 1 ) g(x 1, r) > 0 x 0 x 1 C where g i(x, r) := r i iφ i(x) and is the transpose operator. We can then see that since σ(x + z, r) σ(x, r) z g(x, r), DSC implies that there are diminishing returns along any direction. Theorem 2 (Uniqueness) NE is unique if r > 0 s.t. σ(x, r) := P n i=1 riφi(x) is DSC. The key is that x is a NE if x j maximizes σ(r, x) j, and DSC implies that this maximizer is unique. A sufficient condition is that the matrix [G(x, r) + G (x, r)] is negative definite for x C, where G i,j(x, r) := 2 g(x,r) x i x j. In the bilinear case where φ i(x) = P n j=1 [e ij+x ic ij]x j for e ij R m j and C ij R m i m j, there 2 is a unique NE if every eigenvalue of C has a negative real part (where C := 6 4 2C 1,1 C 1,2... C 1,mj C 2,1 C 2,2... C 2,mj C mi,1 C mi,2... C mi,m j We can get a more intuitive feel for all of this by looking at local improvements. Choose r > 0 and define x(t), t > 0 by dx i = r dt i iφ i(x) + P j uj ihj(x) where the uj s project on C. Then if G + G is negative definite, the unique equilibrium is assymptotically stable. We can see this graphically in figure 2. With appropriate r i s and u j s x(t) stays in C. Eventually, it will move to the Nash Equilibrium. 2 Learning in Games 2.1 Motivation We now return to our simpler games with a discrete set of actions for each player, but extend to repeated games. Through repetition of the game, the players will learn about the strategies of 2
3 Figure 2: Local Improvements the other players. We will then be able to explain equilibria as the result of the players learning over time rather than being fully rational. This is somewhat more intuitively satisfying, because the fully rational assumption does not seem to apply to what we see in real life. 2.2 Examples Figure 3: A 2-player game with a Nash Equilibrium (D,L) strictly dominated by another strategy (U,R). We will look first at a simple 2-player game with a reward matrix given in figure 3. We can see that the Nash Equilibrium is (D,L), but both players can benefit if they play (U,R) instead. If P1 is patient and knows P2 chooses her play based on her forecast of P1s plays, then P1 should always play U to lead P2 to play R. In this way, a sophisticated and patient player who faces a nave opponent can develop a reputation for playing a fixed strategy and obtain the rewards of a Stackelberg leader. However, most theory avoids this situation by assuming random pairings in a large population. Without being able to assume the other player is naïve, the myopic strategy is optimal. It is also important to note that with repeated games, we need to assume the games are repeated infintely many times (or the players do not know how many times they will be played). We can see this through the example game in figure 3 repeated 100 times. Suppose you are P1. You may want to play U to encourage P2 to play R, but on the last trial, there are no future benefits for playing U, so you should switch to D. Knowing that your actions do not affect the future, the myopic strategy is clearly the right way to play. But we can iterate on this result. Now that the strategies are fixed for the 100th trial, actions on the 99th trial do not affect the future. It should therefore also be played NE. By induction, it follows that every trial should be played NE. In the Cournot game, we can reach a Nash Equilibrium without assuming fully rational players with full information. The repeating of the game allows the players to learn, adjusting his strategy to the best response to the strategy of the other player from the previous trial. This adjustment is a bit naïve in that it ignores the effect your previous play has on the other player, but it converges to NE nonetheless. 3
4 2.3 Models A learning model specifies rules of individual players and examines their interactions in repeated game. Usually the same game is repeated, though a few studies have been done on learning from similar games. Three models are: 1. Fictitious Play: Players observe result of their own match, play best response to the historical frequency of play. 2. Partial Best-Response Dynamics: In each period, a fixed fraction of the population switches to a best response to the aggregate statistics from the previous period. 3. Replicator Dynamics: Share of population using each strategy grows at a rate proportional to that strategys current payoff. 2.4 Fictitious Play In fictitious play, each player computes the frequency of the actions of the other players (with initial weights). Each player then selects best response to the empirical distribution. Theorem 3 Strict Nash Equilibria are absorbing for Fictitious Play (If s is a pure strategy and is steady-state for FP, then it is a NE) Proof: Assume s(t) = s is a strict NE. Let a = a(t) be the proportion of the players that are playing s at time t. The strategies played at time t + 1 are The utility for playing strategy r at time t + 1 is then p(t + 1) = (1a)p(t) + aδ(s), (1) u(t + 1, r) = (1a)u(p(t), r) + au(d(s), r), (2) which is maximized by r = s if u(p(t), r) is maximized by r = s. Figure 4: A reward matrix for the matching pennies game We can see an example of convergence in the matching pennies game, for which the reward matrix is given in figure 4. Suppose P1 has initial weights of (1.5,2) and P2 has initial weights of (2,1.5). Then the strategies progress as follows: Strategy New Empirical Distribution (T,T) (1.5,3),(2,2.5) (T,H) (2.5,3),(2,3.5) (T,H) (3.5,3),(2,4.5) (H,H) (4.5,3),(3,4.5) (H,H) (5.5,3),(4,4.5) (H,H) (6.5,3),(5,4.5) (H,T) (6.5,4),(6,4.5) Eventually, the distribution converges to each player playing 50/50. 4
5 Theorem 4 If under FP empirical converge, then product converges to NE Proof: Intuitively, if strategies converge, this means players do not want to deviate, so limit must be NE. Further, there are conditions that guarantee convergence of an empirical FP. These include, but are not limited to: 2 2 games with generic payoffs zero-sum games games solvable by iterated strict dominance Figure 5: A reward matrix for the matching pennies game But not all empirical distributions converge. Take, for example the coordination game with reward matrix given in figure 5, with initial weights (1, 2) for both P1 and P2. Then the strategies progress as follows: Strategy New Empirical Distribution (A,A) (2, 2) (B,B) (2, 1 + 2) (A,A) (3, 1 + 2) (B,B) (3, 2 + 2) The empirical frequencies converge to the Nash Equilibrium, yet the players get 0. The problem is that even though the frequencies are right, the players choices are correlated. They are not independent. This problem can be fixed by adding some randomness to the players strategies. This leads us to stochastic fictitious play. 2.5 Stochastic Fictitious Play The goal of stochastic fixed play is to get a stronger form of convergence not only of the marginals, but also of the intended plays. Let s be the strategies being played. We will define the reward of player i to be u(i, s)+n(i, s i ) where n has positive support on the interval, and s i is the strategies of the players other than player i. The best response on player i to a strategy distribution σ is BR(i, σ)(s i ) = P [n(i, s i )is s.t.s i = BRtoσ] (3) σ is a Nash distribution if σ i = BR(i, σ) i. Harsayni s Purification Theorem states that for generic payoffs, the Nash Distribution approaches the Nash Equilibrium as the support of the perturbation approaches 0. The key feature of this stochistic fictitious play model is that the BR curve is close to the original BR curve but with the discontinuities removed. This is illustrated in figure 6. Other results regarding stochasitc fictitious play has also been found recently. In 1993 Fudenberg and Kreps found that if smoothing is small enough in a 2 2 game with a unique mixed NE, then the NE is globally stable for SFP. In 1995, it was shown that for a 2 2 game with a unique strict NE, the unique intersection of the smoothed BR curve is a global attractor for SFP. 5
6 Figure 6: The best-response curve for the matching pennies game. The red curve is the regular BR curve, while the blue curve includes a random perturbation. Then, in 1996, it was shown that if a 2 2 game has 2 strict NE and one mixed NE, then the SFP converges to one of the strict NE with probability 1. While SFP has all of these nice results, it is important to note that if we increase the number of players, these results do not hold and cycling is still possible. Another possible justification for randomization is as protection against opponents mistakes. A learning rule should be safe (average utility minimax) and universally consistent (utility at least as good as if we knew the frequencies but not the order of plays). The universal consistency can be satisfied by randomization as we saw in the matching pennies example. There is also an alternative to SFP called Stimulus-Response (Reinforcement Learning). In SR, you increase the probability of plays that give good results. It is difficult to discriminate learning models on the basis of experimental data: SFP, SR, etc. seem all about comparable. 3 Cooperative Games 3.1 Motivation We have seen that in some games, like the Cournot game, the NE is not the most desirable outcome for the players. They can do better if they cooperate. In this section, we will In this section, we will explore some notions of equilibrium that players achieve under cooperation. 3.2 Notions of Equilibria Figure 7: Some different equilibria in a 2-player game: The shaded region is dominated by its border which is the pareto set. If we sweep a line with slope -1, the highest point at which the line is entirely in the shaded region gives us the max-min, and the lowest point at which the line is entireley in the unshaded region gives us the social welfare. Figure 7 shows some of the possible points we can operate at. x j is the reward of player j. Max-min is a nice point because it is the best for the player who gets less, but it is not that efficient of an equilibrium. On the other hand, the social equilibrium maximizes the total utility, but this can be very unfair in that one player may get much more that the other. The Nash bargaining equilibrium is a nice compromise. 6
7 3.3 Nash Bargaining Equilibrium Definition 4 (Nash Bargaining Equilibrium (NBE)) x N P x j x N j j x N j 0. is a NBE if x R one has Intuitively, this means that moving to any other vector of rewards results in a negative sum of relative changes of the rewards. The NBE maximizes Q j xj over the feasible vectors of rewards. Let us use an example in which Alice and Bob must decide how to share $100. Alice s utility for x dollars is a(x) = 10 + x. Bob s utility for x dollars is b(x) = x. In the NBE, we want x NBE = arg max x p (10 + x)(80 + 2(100 x)) = arg max x p (10 + x)(280 2x) = arg max x(10 + x)(280 2x) = 65 (4) This means that Alice gets x = 65 dollars and Bob gets the remaining = 35 dollars. Total welfare = a(65) + b(35) = = 20.8 Using the social equilibrium, x social = arg max p x( (10 + x) + p (80 + 2(100 x))) = 40 (5) Total welfare = a(40) + b(60) = = We can see that the NBE has less total utility than the social equilibrium. Figure 8: An axiomatic justification of the Nash Bargaining Equilibrium: In the first figure, both players have equal utility so the point where R 1 = R 2 is best. In the second, some of the points are no longer feasible, but the old fair equilibrium is, so it should still be considered the best. In the third figure, the R 1 axis is stretched, while the R 2 axis is compressed, but since a fair equilibrium should be independent of currency, (a 1 /2, a 2 /2) should still be the best equilibrium Nevertheless, we consider the NBE better in some some sense. justification. This idea can be seen in figure 8. It has a nice axiomatic 3.4 Shapely Value The Shapely Value addresses a slightly different kind of problem. We now want to split the money based on each player s value not their utility. Let us look for example at a situation in which Farmer Bob hires two workers. With one worker, he can produce $ With two he can produce $ Alone he can produce nothing, and without Bob, the workers can produce nothing. What is the fair way to split the $200.00? To determine the Shapely Value of each contributor, we assume they show up in random order. We will call one s contribution the marginal increase in value when he arrives. Each person should then receive a share equal to his average contribution. In this case, Bob s average contribution is $100, and the workers each have an average contribution of $50. 7
8 4 Fixed-Point Theorems 4.1 Brower Theorem 5 (Brower) Let S = {x = (x 1,..., x n) 0 P j xj = 1} and f : S S a continuous function. The function f admits a fixed point. Figure 9: The triangle used for the proof of Brower s theorem For the proof, we draw a large triangle that contains S (figure 9). We then label the corners as shown in figure 9. We divide the big triangle into many small triangles as shown. We then give each vertex a label according to the following procedure: The vertex represents some value x. Draw a ray from the vertex in the direction of f(x). The ray must exit the triangle crossing one of the edges. Label the vertex with the same label as the corner opposite the edge which the ray exits through. Figure 10: A fully labeled triangle and paths through doors A fully labeled triangle is shown in figure 10. Note that the vertices on an edge can only be labeled with the same number as one of the endpoints of that edge. This is becuase f maps S to S, so the rays must point into the triangle. We can now show that there must be a triangle with all three vertices with different labels (call this a triangle). We do this by calling a door any edge that connects a 1 and a 2. We can then draw paths through the triangle that only go through doors. As long as the path does not go into a triangle, there must be a door out of the triangle. But we observe that the bottom edge of the big triangle must have an odd number of doors. This is because there is a 1 on the far left and a 2 on the far right. Since there are no 3 s there must be an odd number of transitions from 1 to 2 and back. This means that at least one path that goes into the triangle cannot come out. This path must end in a triangle. We can then iterate this algorithm on the triangle repeatedly until we have an infinitely small triangle. In the limit, the triangle will converge to a point. Call it z. If f(z) were outside of the triangle, the triangle would not be a triangle (see figure 11). So f(z) must be in the triangle. But in the limit, the triangle is just a point. This means that f(z) = z. z is a fixed point for f. 8
9 Figure 11: An infinitely small triangle containing the point z 4.2 Kakutani Theorem 6 (Kakutani) Let S be a simplex and f( ) : S 2 S a function. Assume that this function is nonempty and convex, which means that f(x) and f(x) is a convex set for all x S. Assume further that this function is graph-continuous. Then there is a fixed point, i.e. some x S with x f(x). Here graph-continuous means that if u n f(s n) for n 1 and s n, u n) (s, u) as n, then u f(s). Proof: Triangulate the simplex. For the nth triangulation, define the function f n as follows: For a vertex x of the triangulation, we pick a point arbitrarily in f(x) and we call that point f n(x). For a point x that is not a vertex of a triangle, we define f n(x) as the linear interpolation of the values of f n at the vertices of the triangle of x. The function f n is continuous and, by Brower s theorem, admits a fixed point, x n. The sequence {x n, n 1} has a limit point x. Using the graph continuity of f( ), we show that x f(x). Designate by v n the triangulation vertex closest to x n, breaking ties arbitrarily. Also, let u n = f n(v n) f(v n). The triangulation is getting finer so that v n x. Also, because of the linear interpolation, u n x. The graph continuity then implies that x f(x). References [1] J.B. Rosen, Existence and Uniqueness of Equilibrium Points for Concave N-Person Games, Econometrica, 33, , July [2] Fudenberg D. and D.K. Levine The Theory of Learning in Games, MIT Press, Cambridge, Massachusetts. Chapters 1, 2, 4,
15-451/651: Design & Analysis of Algorithms October 11, 2018 Lecture #13: Linear Programming I last changed: October 9, 2018
15-451/651: Design & Analysis of Algorithms October 11, 2018 Lecture #13: Linear Programming I last changed: October 9, 2018 In this lecture, we describe a very general problem called linear programming
More informationEC422 Mathematical Economics 2
EC422 Mathematical Economics 2 Chaiyuth Punyasavatsut Chaiyuth Punyasavatust 1 Course materials and evaluation Texts: Dixit, A.K ; Sydsaeter et al. Grading: 40,30,30. OK or not. Resources: ftp://econ.tu.ac.th/class/archan/c
More information6 Extensive Form Games
6 Extensive Form Games 6.1 Example: Representing a Simultaneous 22 Game Alice H HHHH O H HH o Q Bob H QQQ h o HHHH h 2 1 1 2 Figure 1: The Battle of the Sexes in Extensive Form So far we have described
More informationCombinatorial Gems. Po-Shen Loh. June 2009
Combinatorial Gems Po-Shen Loh June 2009 Although this lecture does not contain many offical Olympiad problems, the arguments which are used are all common elements of Olympiad problem solving. Some of
More informationIntroduction to Algorithms / Algorithms I Lecturer: Michael Dinitz Topic: Algorithms and Game Theory Date: 12/3/15
600.363 Introduction to Algorithms / 600.463 Algorithms I Lecturer: Michael Dinitz Topic: Algorithms and Game Theory Date: 12/3/15 25.1 Introduction Today we re going to spend some time discussing game
More informationPascal De Beck-Courcelle. Master in Applied Science. Electrical and Computer Engineering
Study of Multiple Multiagent Reinforcement Learning Algorithms in Grid Games by Pascal De Beck-Courcelle A thesis submitted to the Faculty of Graduate and Postdoctoral Affairs in partial fulfillment of
More informationCONSUMPTION BASICS. MICROECONOMICS Principles and Analysis Frank Cowell. July 2017 Frank Cowell: Consumption Basics 1
CONSUMPTION BASICS MICROECONOMICS Principles and Analysis Frank Cowell July 2017 Frank Cowell: Consumption Basics 1 Overview Consumption: Basics The setting The environment for the basic consumer optimisation
More informationOn a Network Generalization of the Minmax Theorem
On a Network Generalization of the Minmax Theorem Constantinos Daskalakis Christos H. Papadimitriou {costis, christos}@cs.berkeley.edu February 10, 2009 Abstract We consider graphical games in which edges
More informationMIDTERM EXAMINATION Networked Life (NETS 112) November 21, 2013 Prof. Michael Kearns
MIDTERM EXAMINATION Networked Life (NETS 112) November 21, 2013 Prof. Michael Kearns This is a closed-book exam. You should have no material on your desk other than the exam itself and a pencil or pen.
More informationAdvanced Operations Research Techniques IE316. Quiz 1 Review. Dr. Ted Ralphs
Advanced Operations Research Techniques IE316 Quiz 1 Review Dr. Ted Ralphs IE316 Quiz 1 Review 1 Reading for The Quiz Material covered in detail in lecture. 1.1, 1.4, 2.1-2.6, 3.1-3.3, 3.5 Background material
More informationMathematical and Algorithmic Foundations Linear Programming and Matchings
Adavnced Algorithms Lectures Mathematical and Algorithmic Foundations Linear Programming and Matchings Paul G. Spirakis Department of Computer Science University of Patras and Liverpool Paul G. Spirakis
More informationCAP 5993/CAP 4993 Game Theory. Instructor: Sam Ganzfried
CAP 5993/CAP 4993 Game Theory Instructor: Sam Ganzfried sganzfri@cis.fiu.edu 1 Announcements HW 1 due today HW 2 out this week (2/2), due 2/14 2 Definition: A two-player game is a zero-sum game if for
More informationNash equilibria in Voronoi Games on Graphs
Nash equilibria in Voronoi Games on Graphs Christoph Dürr, Nguyễn Kim Thắng (Ecole Polytechnique) ESA, Eilat October 07 Plan Motivation : Study the interaction between selfish agents on Internet k players,
More informationAlgorithmic Game Theory - Introduction, Complexity, Nash
Algorithmic Game Theory - Introduction, Complexity, Nash Branislav Bošanský Czech Technical University in Prague branislav.bosansky@agents.fel.cvut.cz February 25, 2018 About This Course main topics of
More informationOn the Computational Complexity of Nash Equilibria for (0, 1) Bimatrix Games
On the Computational Complexity of Nash Equilibria for (0, 1) Bimatrix Games Bruno Codenotti Daniel Štefankovič Abstract The computational complexity of finding a Nash equilibrium in a nonzero sum bimatrix
More informationExact Algorithms Lecture 7: FPT Hardness and the ETH
Exact Algorithms Lecture 7: FPT Hardness and the ETH February 12, 2016 Lecturer: Michael Lampis 1 Reminder: FPT algorithms Definition 1. A parameterized problem is a function from (χ, k) {0, 1} N to {0,
More informationTHREE LECTURES ON BASIC TOPOLOGY. 1. Basic notions.
THREE LECTURES ON BASIC TOPOLOGY PHILIP FOTH 1. Basic notions. Let X be a set. To make a topological space out of X, one must specify a collection T of subsets of X, which are said to be open subsets of
More informationStrategic Network Formation
Strategic Network Formation Zhongjing Yu, Big Data Research Center, UESTC Email:junmshao@uestc.edu.cn http://staff.uestc.edu.cn/shaojunming What s meaning of Strategic Network Formation? Node : a individual.
More informationSimple Channel-Change Games for Spectrum- Agile Wireless Networks
1 Proceedings of Student/Faculty Research Day, CSIS, Pace University, May 5 th, 26 Simple Channel-Change Games for Spectrum- Agile Wireless Networks Roli G. Wendorf and Howard Blum Abstract The proliferation
More informationComment on Strategic Information Management Under Leakage in a. Supply Chain
Comment on Strategic Information Management Under Leakage in a Supply Chain Lin Tian 1 School of International Business Administration, Shanghai University of Finance and Economics, 00433 Shanghai, China,
More informationSolutions of Stochastic Coalitional Games
Applied Mathematical Sciences, Vol. 8, 2014, no. 169, 8443-8450 HIKARI Ltd, www.m-hikari.com http://dx.doi.org/10.12988/ams.2014.410881 Solutions of Stochastic Coalitional Games Xeniya Grigorieva St.Petersburg
More informationPolygon Partitioning. Lecture03
1 Polygon Partitioning Lecture03 2 History of Triangulation Algorithms 3 Outline Monotone polygon Triangulation of monotone polygon Trapezoidal decomposition Decomposition in monotone mountain Convex decomposition
More informationStochastic Coalitional Games with Constant Matrix of Transition Probabilities
Applied Mathematical Sciences, Vol. 8, 2014, no. 170, 8459-8465 HIKARI Ltd, www.m-hikari.com http://dx.doi.org/10.12988/ams.2014.410891 Stochastic Coalitional Games with Constant Matrix of Transition Probabilities
More information(67686) Mathematical Foundations of AI July 30, Lecture 11
(67686) Mathematical Foundations of AI July 30, 2008 Lecturer: Ariel D. Procaccia Lecture 11 Scribe: Michael Zuckerman and Na ama Zohary 1 Cooperative Games N = {1,...,n} is the set of players (agents).
More informationAlgorithms, Games, and Networks March 28, Lecture 18
Algorithms, Games, and Networks March 28, 2013 Lecturer: Ariel Procaccia Lecture 18 Scribe: Hanzhang Hu 1 Strategyproof Cake Cutting All cake cutting algorithms we have discussed in previous lectures are
More informationLecture Notes on Congestion Games
Lecture Notes on Department of Computer Science RWTH Aachen SS 2005 Definition and Classification Description of n agents share a set of resources E strategy space of player i is S i 2 E latency for resource
More informationA New Approach to Meusnier s Theorem in Game Theory
Applied Mathematical Sciences, Vol. 11, 2017, no. 64, 3163-3170 HIKARI Ltd, www.m-hikari.com https://doi.org/10.12988/ams.2017.712352 A New Approach to Meusnier s Theorem in Game Theory Senay Baydas Yuzuncu
More informationFast Convergence of Regularized Learning in Games
Fast Convergence of Regularized Learning in Games Vasilis Syrgkanis Alekh Agarwal Haipeng Luo Robert Schapire Microsoft Research NYC Microsoft Research NYC Princeton University Microsoft Research NYC Strategic
More informationISM206 Lecture, April 26, 2005 Optimization of Nonlinear Objectives, with Non-Linear Constraints
ISM206 Lecture, April 26, 2005 Optimization of Nonlinear Objectives, with Non-Linear Constraints Instructor: Kevin Ross Scribe: Pritam Roy May 0, 2005 Outline of topics for the lecture We will discuss
More informationToday s lecture. Competitive Matrix Games. Competitive Matrix Games. Modeling games as hybrid systems. EECE 571M/491M, Spring 2007 Lecture 17
EECE 57M/49M, Spring 007 Lecture 7 Modeling games as hybrid systems oday s lecture Background Matrix games Nash Competitive Equilibrium Nash Bargaining Solution Strategy dynamics: he need for hybrid models
More informationSimplicial Hyperbolic Surfaces
Simplicial Hyperbolic Surfaces Talk by Ken Bromberg August 21, 2007 1-Lipschitz Surfaces- In this lecture we will discuss geometrically meaningful ways of mapping a surface S into a hyperbolic manifold
More informationCore Membership Computation for Succinct Representations of Coalitional Games
Core Membership Computation for Succinct Representations of Coalitional Games Xi Alice Gao May 11, 2009 Abstract In this paper, I compare and contrast two formal results on the computational complexity
More informationAlgorithmic Game Theory. Alexander Skopalik
Algorithmic Game Theory Alexander Skopalik Today 1. Strategic games (Normal form games) Complexity 2. Zero Sum Games Existence & Complexity Outlook, next week: Network Creation Games Guest lecture by Andreas
More information1. Lecture notes on bipartite matching February 4th,
1. Lecture notes on bipartite matching February 4th, 2015 6 1.1.1 Hall s Theorem Hall s theorem gives a necessary and sufficient condition for a bipartite graph to have a matching which saturates (or matches)
More informationA Signaling Game Approach to Databases Querying
A Signaling Game Approach to Databases Querying Ben McCamish 1, Arash Termehchy 1, Behrouz Touri 2, and Eduardo Cotilla-Sanchez 1 1 School of Electrical Engineering and Computer Science, Oregon State University
More informationThe Price of Selfishness in Network Coding
The Price of Selfishness in Network Coding Jason R. Marden and Michelle Effros Abstract We introduce a game theoretic framework for studying a restricted form of network coding in a general wireless network.
More informationAlgorithmic Game Theory and Applications. Lecture 16: Selfish Network Routing, Congestion Games, and the Price of Anarchy.
Algorithmic Game Theory and Applications Lecture 16: Selfish Network Routing, Congestion Games, and the Price of Anarchy Kousha Etessami games and the internet Basic idea: The internet is a huge experiment
More informationEC 521 MATHEMATICAL METHODS FOR ECONOMICS. Lecture 2: Convex Sets
EC 51 MATHEMATICAL METHODS FOR ECONOMICS Lecture : Convex Sets Murat YILMAZ Boğaziçi University In this section, we focus on convex sets, separating hyperplane theorems and Farkas Lemma. And as an application
More information2D rendering takes a photo of the 2D scene with a virtual camera that selects an axis aligned rectangle from the scene. The photograph is placed into
2D rendering takes a photo of the 2D scene with a virtual camera that selects an axis aligned rectangle from the scene. The photograph is placed into the viewport of the current application window. A pixel
More informationOn the Structure of Weakly Acyclic Games
1 On the Structure of Weakly Acyclic Games Alex Fabrikant (Princeton CS Google Research) Aaron D. Jaggard (DIMACS, Rutgers Colgate University) Michael Schapira (Yale CS & Berkeley CS Princeton CS) 2 Best-
More informationAlgorithmic Game Theory and Applications. Lecture 16: Selfish Network Routing, Congestion Games, and the Price of Anarchy
Algorithmic Game Theory and Applications Lecture 16: Selfish Network Routing, Congestion Games, and the Price of Anarchy Kousha Etessami warning, again 1 In the few remaining lectures, we will briefly
More information1. Lecture notes on bipartite matching
Massachusetts Institute of Technology 18.453: Combinatorial Optimization Michel X. Goemans February 5, 2017 1. Lecture notes on bipartite matching Matching problems are among the fundamental problems in
More informationMa/CS 6b Class 26: Art Galleries and Politicians
Ma/CS 6b Class 26: Art Galleries and Politicians By Adam Sheffer The Art Gallery Problem Problem. We wish to place security cameras at a gallery, such that they cover it completely. Every camera can cover
More informationLecture 1. 1 Notation
Lecture 1 (The material on mathematical logic is covered in the textbook starting with Chapter 5; however, for the first few lectures, I will be providing some required background topics and will not be
More informationAMATH 383 Lecture Notes Linear Programming
AMATH 8 Lecture Notes Linear Programming Jakob Kotas (jkotas@uw.edu) University of Washington February 4, 014 Based on lecture notes for IND E 51 by Zelda Zabinsky, available from http://courses.washington.edu/inde51/notesindex.htm.
More informationLecture 5: Properties of convex sets
Lecture 5: Properties of convex sets Rajat Mittal IIT Kanpur This week we will see properties of convex sets. These properties make convex sets special and are the reason why convex optimization problems
More informationMultiple Agents. Why can t we all just get along? (Rodney King) CS 3793/5233 Artificial Intelligence Multiple Agents 1
Multiple Agents Why can t we all just get along? (Rodney King) CS 3793/5233 Artificial Intelligence Multiple Agents 1 Assumptions Assumptions Definitions Partially bservable Each agent can act autonomously.
More informationMTAEA Convexity and Quasiconvexity
School of Economics, Australian National University February 19, 2010 Convex Combinations and Convex Sets. Definition. Given any finite collection of points x 1,..., x m R n, a point z R n is said to be
More informationThe Fundamentals of Economic Dynamics and Policy Analyses: Learning through Numerical Examples. Part II. Dynamic General Equilibrium
The Fundamentals of Economic Dynamics and Policy Analyses: Learning through Numerical Examples. Part II. Dynamic General Equilibrium Hiroshi Futamura The objective of this paper is to provide an introductory
More informationLecture 2: August 29, 2018
10-725/36-725: Convex Optimization Fall 2018 Lecturer: Ryan Tibshirani Lecture 2: August 29, 2018 Scribes: Adam Harley Note: LaTeX template courtesy of UC Berkeley EECS dept. Disclaimer: These notes have
More informationA Course in Machine Learning
A Course in Machine Learning Hal Daumé III 13 UNSUPERVISED LEARNING If you have access to labeled training data, you know what to do. This is the supervised setting, in which you have a teacher telling
More informationMath 5593 Linear Programming Lecture Notes
Math 5593 Linear Programming Lecture Notes Unit II: Theory & Foundations (Convex Analysis) University of Colorado Denver, Fall 2013 Topics 1 Convex Sets 1 1.1 Basic Properties (Luenberger-Ye Appendix B.1).........................
More informationLecture 3: Art Gallery Problems and Polygon Triangulation
EECS 396/496: Computational Geometry Fall 2017 Lecture 3: Art Gallery Problems and Polygon Triangulation Lecturer: Huck Bennett In this lecture, we study the problem of guarding an art gallery (specified
More informationLinear Programming Duality and Algorithms
COMPSCI 330: Design and Analysis of Algorithms 4/5/2016 and 4/7/2016 Linear Programming Duality and Algorithms Lecturer: Debmalya Panigrahi Scribe: Tianqi Song 1 Overview In this lecture, we will cover
More informationE-Companion: On Styles in Product Design: An Analysis of US. Design Patents
E-Companion: On Styles in Product Design: An Analysis of US Design Patents 1 PART A: FORMALIZING THE DEFINITION OF STYLES A.1 Styles as categories of designs of similar form Our task involves categorizing
More informationConvexity: an introduction
Convexity: an introduction Geir Dahl CMA, Dept. of Mathematics and Dept. of Informatics University of Oslo 1 / 74 1. Introduction 1. Introduction what is convexity where does it arise main concepts and
More informationError-Correcting Codes
Error-Correcting Codes Michael Mo 10770518 6 February 2016 Abstract An introduction to error-correcting codes will be given by discussing a class of error-correcting codes, called linear block codes. The
More information{ 1} Definitions. 10. Extremal graph theory. Problem definition Paths and cycles Complete subgraphs
Problem definition Paths and cycles Complete subgraphs 10. Extremal graph theory 10.1. Definitions Let us examine the following forbidden subgraph problems: At most how many edges are in a graph of order
More informationSection 1.5 Transformation of Functions
6 Chapter 1 Section 1.5 Transformation of Functions Often when given a problem, we try to model the scenario using mathematics in the form of words, tables, graphs and equations in order to explain or
More informationLecture 2 September 3
EE 381V: Large Scale Optimization Fall 2012 Lecture 2 September 3 Lecturer: Caramanis & Sanghavi Scribe: Hongbo Si, Qiaoyang Ye 2.1 Overview of the last Lecture The focus of the last lecture was to give
More information16.410/413 Principles of Autonomy and Decision Making
16.410/413 Principles of Autonomy and Decision Making Lecture 17: The Simplex Method Emilio Frazzoli Aeronautics and Astronautics Massachusetts Institute of Technology November 10, 2010 Frazzoli (MIT)
More informationAn Introduction to Bilevel Programming
An Introduction to Bilevel Programming Chris Fricke Department of Mathematics and Statistics University of Melbourne Outline What is Bilevel Programming? Origins of Bilevel Programming. Some Properties
More informationPebble Sets in Convex Polygons
2 1 Pebble Sets in Convex Polygons Kevin Iga, Randall Maddox June 15, 2005 Abstract Lukács and András posed the problem of showing the existence of a set of n 2 points in the interior of a convex n-gon
More informationCDG2A/CDZ4A/CDC4A/ MBT4A ELEMENTS OF OPERATIONS RESEARCH. Unit : I - V
CDG2A/CDZ4A/CDC4A/ MBT4A ELEMENTS OF OPERATIONS RESEARCH Unit : I - V UNIT I Introduction Operations Research Meaning and definition. Origin and History Characteristics and Scope Techniques in Operations
More informationTargeting Nominal GDP or Prices: Expectation Dynamics and the Interest Rate Lower Bound
Targeting Nominal GDP or Prices: Expectation Dynamics and the Interest Rate Lower Bound Seppo Honkapohja, Bank of Finland Kaushik Mitra, University of Saint Andrews *Views expressed do not necessarily
More information/633 Introduction to Algorithms Lecturer: Michael Dinitz Topic: Approximation algorithms Date: 11/27/18
601.433/633 Introduction to Algorithms Lecturer: Michael Dinitz Topic: Approximation algorithms Date: 11/27/18 22.1 Introduction We spent the last two lectures proving that for certain problems, we can
More informationBROUWER S FIXED POINT THEOREM. Contents
BROUWER S FIXED POINT THEOREM JASPER DEANTONIO Abstract. In this paper we prove Brouwer s Fixed Point Theorem, which states that for any continuous transformation f : D D of any figure topologically equivalent
More informationModeling and Simulating Social Systems with MATLAB
Modeling and Simulating Social Systems with MATLAB Lecture 7 Game Theory / Agent-Based Modeling Stefano Balietti, Olivia Woolley, Lloyd Sanders, Dirk Helbing Computational Social Science ETH Zürich 02-11-2015
More informationLecture 3. Corner Polyhedron, Intersection Cuts, Maximal Lattice-Free Convex Sets. Tepper School of Business Carnegie Mellon University, Pittsburgh
Lecture 3 Corner Polyhedron, Intersection Cuts, Maximal Lattice-Free Convex Sets Gérard Cornuéjols Tepper School of Business Carnegie Mellon University, Pittsburgh January 2016 Mixed Integer Linear Programming
More informationLecture Notes 2: The Simplex Algorithm
Algorithmic Methods 25/10/2010 Lecture Notes 2: The Simplex Algorithm Professor: Yossi Azar Scribe:Kiril Solovey 1 Introduction In this lecture we will present the Simplex algorithm, finish some unresolved
More informationLecture 6: Graph Properties
Lecture 6: Graph Properties Rajat Mittal IIT Kanpur In this section, we will look at some of the combinatorial properties of graphs. Initially we will discuss independent sets. The bulk of the content
More informationLecture 1: Sperner, Brouwer, Nash. Philippe Bich, PSE and University Paris 1 Pantheon-Sorbonne, France. Lecture 1: Sperner, Brouwer, Nash
.., PSE and University Paris 1 Pantheon-Sorbonne, France. 1. Simplex A n-simplex (or simplex of dimension n) is (x 0,...x n ) = { n i=0 λ ix i : (λ 0,..., λ n ) R n+1 + : n i=0 λ i = 1}, where x 0,...,
More informationModelling competition in demand-based optimization models
Modelling competition in demand-based optimization models Stefano Bortolomiol Virginie Lurkin Michel Bierlaire Transport and Mobility Laboratory (TRANSP-OR) École Polytechnique Fédérale de Lausanne Workshop
More informationNetwork Topology and Equilibrium Existence in Weighted Network Congestion Games
Network Topology and Equilibrium Existence in Weighted Network Congestion Games Igal Milchtaich, Bar-Ilan University August 2010 Abstract. Every finite noncooperative game can be presented as a weighted
More informationUNIT 3 EXPRESSIONS AND EQUATIONS Lesson 3: Creating Quadratic Equations in Two or More Variables
Guided Practice Example 1 Find the y-intercept and vertex of the function f(x) = 2x 2 + x + 3. Determine whether the vertex is a minimum or maximum point on the graph. 1. Determine the y-intercept. The
More informationHow Bad is Selfish Routing?
How Bad is Selfish Routing? Tim Roughgarden and Éva Tardos Presented by Brighten Godfrey 1 Game Theory Two or more players For each player, a set of strategies For each combination of played strategies,
More informationA Game-Theoretic Framework for Congestion Control in General Topology Networks
A Game-Theoretic Framework for Congestion Control in General Topology SYS793 Presentation! By:! Computer Science Department! University of Virginia 1 Outline 2 1 Problem and Motivation! Congestion Control
More information1 Variations of the Traveling Salesman Problem
Stanford University CS26: Optimization Handout 3 Luca Trevisan January, 20 Lecture 3 In which we prove the equivalence of three versions of the Traveling Salesman Problem, we provide a 2-approximate algorithm,
More informationApplied Lagrange Duality for Constrained Optimization
Applied Lagrange Duality for Constrained Optimization Robert M. Freund February 10, 2004 c 2004 Massachusetts Institute of Technology. 1 1 Overview The Practical Importance of Duality Review of Convexity
More informationAlgorithmic Game Theory and Applications. Lecture 6: The Simplex Algorithm
Algorithmic Game Theory and Applications Lecture 6: The Simplex Algorithm Kousha Etessami Recall our example 1 x + y
More informationGraph Theory II. Po-Shen Loh. June edges each. Solution: Spread the n vertices around a circle. Take parallel classes.
Graph Theory II Po-Shen Loh June 009 1 Warm-up 1. Let n be odd. Partition the edge set of K n into n matchings with n 1 edges each. Solution: Spread the n vertices around a circle. Take parallel classes..
More informationA Short SVM (Support Vector Machine) Tutorial
A Short SVM (Support Vector Machine) Tutorial j.p.lewis CGIT Lab / IMSC U. Southern California version 0.zz dec 004 This tutorial assumes you are familiar with linear algebra and equality-constrained optimization/lagrange
More information6 Randomized rounding of semidefinite programs
6 Randomized rounding of semidefinite programs We now turn to a new tool which gives substantially improved performance guarantees for some problems We now show how nonlinear programming relaxations can
More informationCS261: Problem Set #2
CS261: Problem Set #2 Due by 11:59 PM on Tuesday, February 9, 2016 Instructions: (1) Form a group of 1-3 students. You should turn in only one write-up for your entire group. (2) Submission instructions:
More informationCourse Summary Homework
Course Summary Homework (Max useful score: 100 - Available points: 210) 15-382: Collective Intelligence (Spring 2018) OUT: April 21, 2018, at 1:00am DUE: May 1, 2018 at 1pm - Available late days: 0 Instructions
More informationNOTATION AND TERMINOLOGY
15.053x, Optimization Methods in Business Analytics Fall, 2016 October 4, 2016 A glossary of notation and terms used in 15.053x Weeks 1, 2, 3, 4 and 5. (The most recent week's terms are in blue). NOTATION
More information11.1 Facility Location
CS787: Advanced Algorithms Scribe: Amanda Burton, Leah Kluegel Lecturer: Shuchi Chawla Topic: Facility Location ctd., Linear Programming Date: October 8, 2007 Today we conclude the discussion of local
More informationSection 1.5 Transformation of Functions
Section 1.5 Transformation of Functions 61 Section 1.5 Transformation of Functions Often when given a problem, we try to model the scenario using mathematics in the form of words, tables, graphs and equations
More informationarxiv: v1 [math.lo] 9 Mar 2011
arxiv:03.776v [math.lo] 9 Mar 0 CONSTRUCTIVE PROOF OF BROUWER S FIXED POINT THEOREM FOR SEQUENTIALLY LOCALLY NON-CONSTANT FUNCTIONS BY SPERNER S LEMMA YASUHITO TANAKA Abstract. In this paper using Sperner
More informationConvex sets and convex functions
Convex sets and convex functions Convex optimization problems Convex sets and their examples Separating and supporting hyperplanes Projections on convex sets Convex functions, conjugate functions ECE 602,
More informationNetwork Improvement for Equilibrium Routing
Network Improvement for Equilibrium Routing UMANG BHASKAR University of Waterloo and KATRINA LIGETT California Institute of Technology Routing games are frequently used to model the behavior of traffic
More informationPlanar Graphs. 1 Graphs and maps. 1.1 Planarity and duality
Planar Graphs In the first half of this book, we consider mostly planar graphs and their geometric representations, mostly in the plane. We start with a survey of basic results on planar graphs. This chapter
More informationSAMPLING AND THE MOMENT TECHNIQUE. By Sveta Oksen
SAMPLING AND THE MOMENT TECHNIQUE By Sveta Oksen Overview - Vertical decomposition - Construction - Running time analysis - The bounded moments theorem - General settings - The sampling model - The exponential
More informationWeek 5. Convex Optimization
Week 5. Convex Optimization Lecturer: Prof. Santosh Vempala Scribe: Xin Wang, Zihao Li Feb. 9 and, 206 Week 5. Convex Optimization. The convex optimization formulation A general optimization problem is
More informationEXTREME POINTS AND AFFINE EQUIVALENCE
EXTREME POINTS AND AFFINE EQUIVALENCE The purpose of this note is to use the notions of extreme points and affine transformations which are studied in the file affine-convex.pdf to prove that certain standard
More informationIEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 55, NO. 5, MAY
IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL 55, NO 5, MAY 2007 1911 Game Theoretic Cross-Layer Transmission Policies in Multipacket Reception Wireless Networks Minh Hanh Ngo, Student Member, IEEE, and
More informationEquilibrium Tracing in Bimatrix Games
Equilibrium Tracing in Bimatrix Games Anne Balthasar Department of Mathematics, London School of Economics, Houghton St, London WCA AE, United Kingdom A.V.Balthasar@lse.ac.uk Abstract. We analyze the relations
More informationDistributed Planning in Stochastic Games with Communication
Distributed Planning in Stochastic Games with Communication Andriy Burkov and Brahim Chaib-draa DAMAS Laboratory Laval University G1K 7P4, Quebec, Canada {burkov,chaib}@damas.ift.ulaval.ca Abstract This
More information15-780: MarkovDecisionProcesses
15-780: MarkovDecisionProcesses J. Zico Kolter Feburary 29, 2016 1 Outline Introduction Formal definition Value iteration Policy iteration Linear programming for MDPs 2 1988 Judea Pearl publishes Probabilistic
More informationDiscrete Optimization. Lecture Notes 2
Discrete Optimization. Lecture Notes 2 Disjunctive Constraints Defining variables and formulating linear constraints can be straightforward or more sophisticated, depending on the problem structure. The
More information