1 Sealed Bid Multi-object Auctions with Necessary Bundles and its Application to Spectrum Auctions ver. 1.0 University of Tokyo 東京大学 松井知己 Tomomi Matsui.

Slides:



Advertisements
Similar presentations
Combinatorial Auction
Advertisements

Cognitive Radio Communications and Networks: Principles and Practice By A. M. Wyglinski, M. Nekovee, Y. T. Hou (Elsevier, December 2009) 1 Chapter 17 Auction-based.
1 Outline relationship among topics secrets LP with upper bounds by Simplex method basic feasible solution (BFS) by Simplex method for bounded variables.
Testing Linear Pricing Algorithms for use in Ascending Combinatorial Auctions (A5) Giro Cavallo David Johnson Emrah Kostem.
Chapter 17: Making Complex Decisions April 1, 2004.
Truthful Spectrum Auction Design for Secondary Networks Yuefei Zhu ∗, Baochun Li ∗ and Zongpeng Li † ∗ Electrical and Computer Engineering, University.
Chapter 11 Limitations of Algorithm Power Copyright © 2007 Pearson Addison-Wesley. All rights reserved.
CPS Bayesian games and their use in auctions Vincent Conitzer
Continuation Methods for Structured Games Ben Blum Christian Shelton Daphne Koller Stanford University.
This Segment: Computational game theory Lecture 1: Game representations, solution concepts and complexity Tuomas Sandholm Computer Science Department Carnegie.
Combinatorial auctions Vincent Conitzer v( ) = $500 v( ) = $700.
Multi-item auctions with identical items limited supply: M items (M smaller than number of bidders, n). Three possible bidder types: –Unit-demand bidders.
Game Theoretical Insights in Strategic Patrolling: Model and Analysis Nicola Gatti – DEI, Politecnico di Milano, Piazza Leonardo.
Preference Elicitation Partial-revelation VCG mechanism for Combinatorial Auctions and Eliciting Non-price Preferences in Combinatorial Auctions.
Seminar In Game Theory Algorithms, TAU, Agenda  Introduction  Computational Complexity  Incentive Compatible Mechanism  LP Relaxation & Walrasian.
Introduction to Game Theory
What is game theory… Game theory studies settings where multiple parties (agents) each have –different preferences (utility functions), –different actions.
CS774. Markov Random Field : Theory and Application Lecture 17 Kyomin Jung KAIST Nov
Algorithmic Game Theory - Basic Solution Concepts and Computational Issues Éva Tardos and Vijay V. Vazirani Presentation Reiknirit, rökfræði og reiknanleiki.
Interchanging distance and capacity in probabilistic mappings Uriel Feige Weizmann Institute.
Algorithmic Applications of Game Theory Lecture 8 1.
Game-Theoretic Approaches to Multi-Agent Systems Bernhard Nebel.
Welfare Maximization in Congestion Games Liad Blumrosen and Shahar Dobzinski The Hebrew University.
1 Computing Nash Equilibrium Presenter: Yishay Mansour.
NP-Complete Problems Reading Material: Chapter 10 Sections 1, 2, 3, and 4 only.
An Introduction to Game Theory Part III: Strictly Competitive Games Bernhard Nebel.
Combinatorial Auction. Conbinatorial auction t 1 =20 t 2 =15 t 3 =6 f(t): the set X  F with the highest total value the mechanism decides the set of.
Convergence Time to Nash Equilibria in Load Balancing Eyal Even-Dar, Tel-Aviv University Alex Kesselman, Tel-Aviv University Yishay Mansour, Tel-Aviv University.
Nash Equilibria in Competitive Societies Eyal Rozenberg Roy Fox.
Chapter 11: Limitations of Algorithmic Power
Algorithms and Economics of Networks Abraham Flaxman and Vahab Mirrokni, Microsoft Research.
Distributed Combinatorial Optimization
Competitive Auctions and Digital Goods Andrew Goldberg, Jason Hartline, and Andrew Wright presenting: Keren Horowitz, Ziv Yirmeyahu.
Multi-item auctions & exchanges (multiple distinguishable items for sale) Tuomas Sandholm Carnegie Mellon University.
The Weighted Proportional Allocation Mechanism Milan Vojnović Microsoft Research Joint work with Thành Nguyen Harvard University, Nov 3, 2009.
CPS 296.1: Computational Microeconomics: Game Theory, Social Choice, and Mechanism Design Instructor: Vincent Conitzer (Assistant Professor of Computer.
Yang Cai Sep 8, An overview of the class Broad View: Mechanism Design and Auctions First Price Auction Second Price/Vickrey Auction Case Study:
1.3 Modeling with exponentially many constr.  Some strong formulations (or even formulation itself) may involve exponentially many constraints (cutting.
The Multiplicative Weights Update Method Based on Arora, Hazan & Kale (2005) Mashor Housh Oded Cats Advanced simulation methods Prof. Rubinstein.
Introduction 1 TexPoint fonts used in EMF. Read the TexPoint manual before you delete this box.: AAA A.
The Power of the Defender M. Gelastou  M. Mavronicolas  V. Papadopoulou  A. Philippou  P. Spirakis §  University of Cyprus, Cyprus § University of.
Presenter: Jen Hua Chi Adviser: Yeong Sung Lin Network Games with Many Attackers and Defenders.
Combinatorial Auctions By: Shai Roitman
Combinatorial Auctions ( Bidding and Allocation) Adapted from Noam Nisan.
Automated Design of Multistage Mechanisms Tuomas Sandholm (Carnegie Mellon) Vincent Conitzer (Carnegie Mellon) Craig Boutilier (Toronto)
Yang Cai Oct 08, An overview of today’s class Basic LP Formulation for Multiple Bidders Succinct LP: Reduced Form of an Auction The Structure of.
How does the market of sponsored links operate? User enters a query The auction for the link to appear on the search results page takes place Advertisements.
OPTIMUM INTEGRATED LINK SCHEDULING AND POWER CONTROL FOR MULTI-HOP WIRELESS NETWORKS Arash Behzad, and Izhak Rubin, IEEE Transactions on Vehicular Technology,
A Study of Central Auction Based Wholesale Electricity Markets S. Ceppi and N. Gatti.
Implicit Hitting Set Problems Richard M. Karp Erick Moreno Centeno DIMACS 20 th Anniversary.
6.853: Topics in Algorithmic Game Theory Fall 2011 Constantinos Daskalakis Lecture 22.
Variations of the Prize- Collecting Steiner Tree Problem Olena Chapovska and Abraham P. Punnen Networks 2006 Reporter: Cheng-Chung Li 2006/08/28.
Auctions serve the dual purpose of eliciting preferences and allocating resources between competing uses. A less fundamental but more practical reason.
Combinatorial Auction. A single item auction t 1 =10 t 2 =12 t 3 =7 r 1 =11 r 2 =10 Social-choice function: the winner should be the guy having in mind.
1 Quality of Routing Congestion Games in Wireless Sensor Networks Costas Busch Louisiana State University Rajgopal Kannan Louisiana State University Athanasios.
TU/e Algorithms (2IL15) – Lecture 12 1 Linear Programming.
Network Formation Games. NFGs model distinct ways in which selfish agents might create and evaluate networks We’ll see two models: Global Connection Game.
The NP class. NP-completeness Lecture2. The NP-class The NP class is a class that contains all the problems that can be decided by a Non-Deterministic.
TU/e Algorithms (2IL15) – Lecture 12 1 Linear Programming.
Advanced Subjects in GT Prepared by Rina Talisman Introduction Revenue Equivalence The Optimal Auction (Myerson 1981) Auctions.
Comp/Math 553: Algorithmic Game Theory Lecture 10
Chapter 10 NP-Complete Problems.
Bayesian games and their use in auctions
Non-additive Security Games
1.3 Modeling with exponentially many constr.
Competitive Auctions and Digital Goods
Chapter 11 Limitations of Algorithm Power
Flow Feasibility Problems
CPS Bayesian games and their use in auctions
Presentation transcript:

1 Sealed Bid Multi-object Auctions with Necessary Bundles and its Application to Spectrum Auctions ver. 1.0 University of Tokyo 東京大学 松井知己 Tomomi Matsui Iwate Prefectural University 岩手県立大学 渡辺隆裕 Takahiro Watanabe

2 Multi object Auction Multi-object Auction: trading oil leases, furniture, pollution rights, airport time slots, spectrum licenses, and delivery routes, etc. Bidders’ preferences are defined on sets of objects. (combinatorial auction, simultaneous auction) Results: (1) Analysis from the point of view of game theory. (2) Apply the result to spectrum auction.

3 Main result We introduce following assumptions; (1)each bidder has a positive reservation value only for one special subset of objects (necessary bundle) (2)admissible bid is a pair of one subset of objects and its price, Game theoretic approach: show the existence of a Nash equilibrium when bidding unit is sufficiently small Application to spectrum auction: polynomial algorithm for the problem to maximize auctioneer’s revenue explicit description of a Nash equilibrium

4 Purpose of this talk purpose of this talk = Find friends ! Combinatorial optimization Game theory Multi-object Auction Multi-agent system searched on internet ⇒ found PRIMA2001 ⇒ submitted paper ⇒ give a talk ⇒ find friends ⇒ further work ! Communication

5 Definitions (bidders) Game theoretic descriptions N ={1,2,…, n}: players (bidders) M = {1,2,…, m}: objects bundle: subset of objects sealed bid auction: submit bids simultaneously open bid auction (English, Japanese, Dutch,…) strategies (admissible bids) of player i: (B i, b i ) ∈ 2 M ×R + : (bundle, bidding price) The bidding price b i is the amount of money that player i is willing to pay for bundle B i.

6 Assumption (strategies) Assumption 1 each player i submits only one pair of bundle and its price (B i, b i ) ∈ 2 M ×R + : restricted but practical combinatorial auction: each player i submits bidding prices of all the bundles f i : 2 M →R + general but impractical (2 M is a huge family)

7 bidding unit ε : bidding unit (bidding grid) Each bidding price is a non-negative multiple of ε. Z ε ={ε j |j is a non-negative integer} strategies (admissible bids) of player i: (B i, b i ) ∈ 2 M ×Z ε profile of bids : vector of bids of all the players ((B 1,b 1 ),(B 2,b 2 ),…, (B n, b n ))=(B, b) B =(B 1, B 2,…, B n ) b = ( b 1, b 2,…, b n )

8 Definitions (auctioneer) Given a profile of bids ((B 1,b 1 ),(B 2,b 2 ),…, (B n, b n ))=(B, b), auctioneer determines the set of winners which maximizes auctioneer’s revenue. Bundle Assignment Problem BAP(B, b) (winner determination problem) max. b T x = b 1 x 1 + b 2 x 2 + ‥+ b n x n s. t. Ax ≦ 1, x ∈ {0,1} N. A=(a ji ) 0-1matrix {0,1} M×N a ji =1 ⇔ object j is in bundle B i

9 Bundle assignment problem BAP(B, b) max. b T x = b 1 x 1 + b 2 x 2 + ‥+ b n x n s. t. Ax ≦ 1, x ∈ {0,1} N. ● A=(a ji ) 0-1matrix {0,1} M×N a ji =1 ⇔ object j is in bundle B i ● x i =1 ⇔ auctioneer assigns bundle B i to player i. ● Ax ≦ 1: each object must belong to at most one player

10 Bundle assignment problem Bundle assignment problem has many names as winner determination problem, max. weight set packing problem, max. weight independent set problem, and max. weight clique problem. theoretically hard: NP-hard practically tractable: many commercial codes solve BAP efficiently (e.g. CPLEX) [Andersson, Tenhunen and Ygge (2000)]

11 multiple-optimal solutions If BAP has multiple-optimal solutions, then auctioneer chooses an optimal solution uniformly at random. Further work: Construct an algorithm for selecting an optimal solution of BAP uniformly at random. The problem is much harder than BAP.

12 Definitions (bidders) V i (S): Each player i has a non-negative reservation value V i (S) for each bundle S. V i : 2 M → Z δ (non-negative multiple of δ) Assumption 2: Each bidder has a positive reservation value only for one special bundle. ⇒ necessary bundle necessary (bundle, price) of player i : ( T i, v i ) V i (S) = v i ⇔ ( S ⊇ T i ) V i (S) = 0 ⇔ ( otherwise )

13 Nash equilibrium profile of bids : ((B 1,b 1 ),(B 2,b 2 ),…, (B n, b n ))=(B, b) Utility of player i : U i (B, b) U i (B, b)=(V i ( B i ) ー b i ) Pr[player i is selected] profile (B*, b*) is a Nash equilibrium ⇔ ∀ i ∈ N, ∀ (B i, b i ) ∈ 2 M ×Z ε, U i (B*, b*) ≧ U i ((B* -i, b* -i ), (B i, b i )) ((B* -i, b* -i ), (B i, b i )) : profile obtained from (B*, b*) by replacing strategy of player i with (B i, b i )

14 Main results Theorem 2: If the bidding unit ε is sufficiently small, then Nash equilibrium exists. size of bidding unit ε ≦ δ(n2 n +1) δ : unit of reservation value, n: number of players profile (B*, b*) is a Nash equilibrium ⇔ ∀ i ∈ N, ∀ (B i, b i ) ∈ 2 M ×Z ε, U i (B*, b*) ≧ U i ((B* -i, b* -i ), (B i, b i )) ((B* -i, b* -i ), (B i, b i )) : profile obtained from (B*, b*) by replacing strategy of player i with (B i, b i )

15 description of a Nash equilibrium Classify the bidders Passed bidders: bidders contained in every optimal solution of BAP(B, b). Questionable bidders: bidders contained in not all but at least one optimal solution of BAP(B, b). Rejected bidders: bidders never appearing in any optimal solutions of BAP(B, b). Optimal solution set: Ω (B, b): set of all the optimal solutions of BAP(B, b).

16 Nash equilibrium Theorem 1: Following profile (B*, b*) is a Nash equilibrium; questionable bidder i : (B* i, b* i ) = (T i, v i ) rejected bidder i : (B* i, b* i ) = (T i, v i ) passed bidder i : B* i =T i, b* i : minimal vector in Z ε satisfying Ω (B*, b*) = Ω (T, v) (solution sets are equivalent) (T i, v i ): ( necessary bundle, reservation value )

17 Application to spectrum auctions

18 Spectrum auction Spectrum auction: objects: spectrums (frequency channel for cellular phone) are arranged in linear order necessary bundles (T i ): sequences of consecutive channels channels : T2T2 T1T1 T4T4 T3T3

19 Longest path problem BAP corresponding to spectrum auction satisfies the conditions; (1)coefficient matrix A is totally unimodular, (2)liner relaxation of BAP has an integer valued optimal solution, (3)equivalent to longest path problem.

20 longest path problem directed graph T2T2 T1T1 T4T4 T3T3 v2v2 v1v1 v3v3 v4v4 nodes: barrier of channels arcs: ( j, j+1), necessary bundles arc weight = reservation value

21 longest path problem T2T2 T1T1 T4T4 T3T3 v2v2 v1v1 v3v3 v4v4

22 longest path problem T2T2 T1T1 T4T4 T3T3 v2v2 v1v1 v3v3 v4v4 Finding a longest path = winner determination

23 random selection Generally, solving BAP and random selection from multiple-optimal solutions are hard. Spectrum auction: bundle assignment ⇒ longest path problem random selection ⇒ random path generation Key idea: BAP(B, b) ⇒ linear relaxation ⇒ dual problem bundle assignment: dynamic programming random selection: path counting algorithm explicit description of a Nash equilibrium: complementality slackness theorem for linear programming problems (detail is omitted)

24 conclusion Assumption 1 (multi-object auction) each player i submits only one pair of bundle and its price (B i, b i ) ∈ 2 M ×R + Assumption 2: Each bidder has a positive reservation value only for one special bundle, called necessary bundle. Theorem 2: If the bidding unit ε is sufficiently small, then Nash equilibrium exists. Theorem 1 : (Characterization of a Nash equilibrium)

25 END

26 Main results Theorem 2: If the bidding unit ε is sufficiently small, then (pure strategy) Nash equilibrium exists. mixed strategy: Nash showed that every strategic form n-persons game with finite number of strategies has a mixed strategy Nash equilibrium. size of bidding unit ε ≦ δ(n2 n +1) δ : unit of reservation value, n: number of players

27 random selection BAP(B, b) max { b T x | Ax ≦ 1, x ∈ {0,1} N } linear relaxation max{b T x | Ax ≦ 1, x ≧ 0} dual problem min{y T x | y T A ≧ b, y ≧ 0 } y* : optimal dual solution M*={j ∈ M | y* T a i =b i } a i : i th column vector Lemma: M* is the set of passed and questionable bidders. Ordinary dynamic programming procedure ⇒ random selection of longest paths