Download to read offline
![15
Braess’s Paradox
Initial Network: Augmented Network:
All traffic incurs more cost![Braess 68]
• also has physical analogs [Cohen/Horowitz 91]
s t
x 1
½
x1
½
½
½
Cost = 1.5 Cost = 2
s t
x 1
x1
0](https://image.slidesharecdn.com/introductiontoalgorithmicaspectofauctiontheory-140510114419-phpapp01-150607173330-lva1-app6892/75/Introduction-to-algorithmic-aspect-of-auction-theory-15-2048.jpg)
![16
Inefficiency of Nash Flows
Note: selfish routing does not minimize average
delay (observed informally by [Pigou 1920])
• Cost of equilibrium flow = 1•1 + 0•1 = 1
• Cost of optimal (min-cost) flow = ½•½ +½•1 =
¾
s t
x
1
0
1 ½
½](https://image.slidesharecdn.com/introductiontoalgorithmicaspectofauctiontheory-140510114419-phpapp01-150607173330-lva1-app6892/75/Introduction-to-algorithmic-aspect-of-auction-theory-16-2048.jpg)
![15
Braess’s Paradox
Initial Network: Augmented Network:
All traffic incurs more cost![Braess 68]
• also has physical analogs [Cohen/Horowitz 91]
s t
x 1
½
x1
½
½
½
Cost = 1.5 Cost = 2
s t
x 1
x1
0](https://crownmelresort.com/image.slidesharecdn.com/introductiontoalgorithmicaspectofauctiontheory-140510114419-phpapp01-150607173330-lva1-app6892/75/Introduction-to-algorithmic-aspect-of-auction-theory-15-2048.jpg)
![16
Inefficiency of Nash Flows
Note: selfish routing does not minimize average
delay (observed informally by [Pigou 1920])
• Cost of equilibrium flow = 1•1 + 0•1 = 1
• Cost of optimal (min-cost) flow = ½•½ +½•1 =
¾
s t
x
1
0
1 ½
½](https://crownmelresort.com/image.slidesharecdn.com/introductiontoalgorithmicaspectofauctiontheory-140510114419-phpapp01-150607173330-lva1-app6892/75/Introduction-to-algorithmic-aspect-of-auction-theory-16-2048.jpg)
Uploaded byAbner Chih Yi Huang
Introduction to algorithmic aspect of auction theory
This document provides an introduction to algorithmic aspects of auction theory. It discusses connections between computer science and economics, algorithmic game theory, and algorithmic mechanism design. It provides examples of applications like load balancing, routing, and the winner determination problem in combinatorial auctions. Approximation algorithms and truthful mechanisms are important areas of research at the intersection of algorithms and auction theory.
More Related Content
![Digital Signal Processing[ECEG-3171]-Ch1_L07](https://cdn.slidesharecdn.com/ss_thumbnails/dspl7ch3-180427094425-thumbnail.jpg?width=640&height=640&fit=bounds)
Similar to Introduction to algorithmic aspect of auction theory
Introduction to algorithmic aspect of auction theory
- 1. Introduction to Algorithmic Aspectof Auction Theory 2010-01-08 Abner Huang 1 CSBB Lab, Dept. of CS, NTHU.
- 2. Outline • Connections betweenComputer Science and Economics • Algorithmic Game Theory • Algorithmic Mechanism Design 2
- 3. Herbert Alexander Simon •Herbert Alexander Simon (June 15, 1916– February 9, 2001) was an American political scientist, economist, and psychologist. • Notable awards – Turing Award 1975 – Nobel Prize in Economics 1978 3
- 4. Thomas Crombie Schelling ThomasCrombie Schelling (born 14 April 1921) was awarded the 2005 Nobel Memorial Prize in Economic Sciences (shared with Robert Aumann) for "having enhanced our understanding of conflict and cooperation through game-theory analysis." 4
- 5. Micromotives and macrobehavior •Schelling analyzed apartheid with cellular automata, which is invented by John von Neumann and is well-studied by computer scientists. 5
- 6.
- 7. Market Equilibrium • Startingin the 60’s, economists have attempted to use the Arrow-Debreu model and the general equilibrium theory to realistically model actual economies, with the goal of evaluating alternate policy options. – Scarf’s algorithm, – Homotopy methods, – Global Newton’s method 7
- 8. Market Equilibrium • Thereare several algorithmic results in the last 6 years, i.e., PTAS, FPTAS, and polynomial time algorithm based on diophantine approximation and ellipsoid algorithm. • Fisher’s Model with Linear Utility Functions is in P. • Arrow-Debreu’s Model with Linear Utility Functions is in P. 8
- 9.
- 10. Outline • Connections betweenComputer Science and Economics • Algorithmic Game Theory • Algorithmic Mechanism Design 10
- 11.
- 12.
- 13. 13 Braess’s Paradox Initial Network: st x 1 ½ x1 ½ ½ ½ Cost = 1.5
- 14. 14 Braess’s Paradox Initial Network:Augmented Network: s t x 1 ½ x1 ½ ½ ½ Cost = 1.5 Cost = 2 s t x 1 x1 0
- 15. 15 Braess’s Paradox Initial Network:Augmented Network: All traffic incurs more cost![Braess 68] • also has physical analogs [Cohen/Horowitz 91] s t x 1 ½ x1 ½ ½ ½ Cost = 1.5 Cost = 2 s t x 1 x1 0
- 16. 16 Inefficiency of NashFlows Note: selfish routing does not minimize average delay (observed informally by [Pigou 1920]) • Cost of equilibrium flow = 1•1 + 0•1 = 1 • Cost of optimal (min-cost) flow = ½•½ +½•1 = ¾ s t x 1 0 1 ½ ½
- 17. 17 Performance Guarantees Good news:in theoretical CS, have lots of techniques for measuring inefficiency. • motivated by NP-completeness, real-time algorithms, etc. Definition: approximation ratio (w.r.t. some objective function): optimal obj fn value protagonists's obj fn value the closer to 1 the better Price of anarchy := equilibrium/OPT ratio
- 18. Market Equilibrium • Fisher’smodel 18
- 19.
- 20.
- 21.
- 22.
- 23.
- 24.
- 25. Arrow-Debreu model An Auction-BasedAlgorithm Garg et al. proposed an auction-based PTAS for the linear case of the Arrow-Debreu model. 25
- 26.
- 27. Outline • Connections betweenComputer Science and Economics • Algorithmic Game Theory • Algorithmic Mechanism Design 27
- 28. Algorithmic mechanism design •Algorithmic mechanism design (AMD) lies at the intersection of economic game theory and computer science. • Noam Nisan and Amir Ronen, from the Hebrew University of Jerusalem, first coined "Algorithmic mechanism design" Games and Economic Behavior (35): 166–196. 28
- 29. Algorithmic mechanism design •It typically employs the analytic tools of theoretical computer science, such as worst case analysis and approximation ratios, in contrast to classical mechanism design in economics which often makes distributional assumptions about the agents. • It also considers computational constraints to be of central importance: mechanisms that cannot be efficiently implemented in polynomial time are not considered to be viable solutions to a mechanism design problem 29
- 30. 30 Worst-Case Time Complexity n=1000，Time cost for different computer time Time 1 1000 steps/s Time 2 2000 steps/s Time 3 4000 steps/s Time 4 8000 steps/s log2n 0.010 0.005 0.003 0.001 n 1 0.5 0.25 0.125 nlog2n 10 5 2.5 1.25 n1.5 32 16 8 4 n2 1,000 500 250 125 n3 1,000,000 500,000 250,000 125,000 1.1n 1039 1039 1038 1038
- 31. P vs. NP-Hard •P = the set of problems which can be solved in polynomial time ,e.g., – 50*n0.2,3*n+2000, 0.01*n+19861234, 0.001*n80 – Finding maximum of a set of integers is in P. (linear time in fact.) • EXP = the set of problems which can be solved in exponential time. • NP-Hard = the set of problems which might be able to be solved in polynomial time (at least). 31
- 32. Distributed artificial intelligence •Mainstreams in DAI research included the following: – Parallel problem solving: mainly deals with how classic AI concepts can be modified, so that multiprocessor systems and clusters of computers can be used to speed up calculation. – Distributed problem solving (DPS): the concept of agent, autonomous entities that can communicate with each other, was developed to serve as an abstraction for developing DPS systems. – Multi-Agent Based Simulation (MABS): a branch of DAI that builds the foundation for simulations that need to analyze not only phenomena at macro level but also at micro level, as it is in many social simulation scenarios. 32
- 33. Multi-Agent Based Simulation •Too crowded or too desolate in PUB? 33
- 34. Applications of Algorithmicmechanism design • Load balancing: The aggregate power of all computers on the Internet is huge. In a “dream world” this aggregate power will be optimally allocated online among all connected processors. One could imagine CPU-intensive jobs automatically migrating to CPU-servers, caching automatically done by computers with free disk space, etc. Access to data, communication lines and even physical attachments (such as printers) could all be allocated across the Internet. tightly linked systems, and is addressed, in various forms and with varying 34
- 35. Applications of Algorithmicmechanism design • Routing: When one computer wishes to send information to another, the data usually gets routed through various intermediate routers. So far this has been done voluntarily, probably due to the low marginal cost of forwarding a packet. However, when communication of larger amounts of data becomes common (e.g. video), and bandwidth needs to be reserved under various quality of service (QoS) protocols, this altruistic behavior of the routers may no longer hold. If so, we will have to design protocols specifically taking the routers’ self-interest into account. 35
- 36. 36 Ticket Allocation Game Projectlead (Ticket Allocator) (rational and intelligent) Maintenance Engineers (rational and intelligent) effort, time
- 37. 37 Resource Allocation inGrid Computing
- 38. 38 ? Incentive Compatible Broadcastin Ad hoc Wireless Networks
- 39. 39 Tier 1 Tier 2 Tier3 Tier 1: UU Net, Sprint, AT&T, Genuity Tier 2: Regional/National ISPs Tier 3: Residential/Company ISP Internet Routing
- 40.
- 41. Example: Shortest Path Goal:route packets along the lowest-cost path from S to T.
- 42. Example: Shortest Path •Each edge is an agent • People want to send messages to other people
- 43. Winner Determination Problem •Winner Determination Problem – Agents want to bid some goods with all-or-nothing style, and will report their private values. • Maximum Independent Set Problem – Given undirected graph G = (V,E), MIS is to find a maximum independent set X ⊆ V in G. A subset of vertices Y ⊆ V is independent if ∀u, v ∈ Y, (u, v) not in E 43
- 44. Maximum Independent SetProblem
- 45. • Theorem Winnerdetermination (WD) problem is NP-hard. 45
- 46. In general, wecan’t approximate VCG-based mechanisms • Inapproximability for VCG-Based Combinatorial Auctions – By Dave Buchfuhrer et al. To appear in the proceedings of SODA 10. 46
- 47. Can we copewith it? • In general, it is hardly possible. • Some ways: – Solve some special cases. – Approximate some special cases. – Change problem formulations. 47
- 48. The Task AllocationProblem
- 49. The Task AllocationProblem • The MinWork Mechanism The idea is to minimize the total work done. This is no very good solution since our agents are able to work in parallel. – Each task is allocated to the agent who is capable of doing it in a minimal amount of time. – Each agent is given payment equal to the time of the second best agent for every task:
- 50. The Task AllocationProblem • Theorem: MinWork is a truthful n- approximation mechanism. • Proof:
- 51. The Task AllocationProblem • Proof (continue): MinWork belongs to the VGC family (⇒ is truthful)
- 52. VGC mechanism • Anoutput function of a VGC mechanism is required to maximize the objective function. In many cases (e.g. combinatorial auctions (Harstad, Rothkopf and Pekec (1995))), this makes the mechanism computationally intractable. Replacing the optimal algorithm with a non-optimal approximation usually leads to untruthful mechanisms 52
- 53. VGC mechanism • Thus,a VGC mechanism essentially provides a solution for any utilitarian problem (except for the possible problem that there might be dominant strategies other than truth-telling). utilitarian problems (Green and Laffont (It is known that (under mild assumptions) VGC mechanisms are the only truthful implementations for utilitarian problems (Green and Laffont (1977)).for 1977)). 53
- 54.
- 55.
- 56. Online Matching forAdwords Auctions