Download presentation

Presentation is loading. Please wait.

Published byClayton Sandborn Modified over 2 years ago

1
Pure and Bayes-Nash Price of Anarchy for GSP Renato Paes Leme Éva Tardos CornellCornell & MSR

2
Keyword Auctions organic search results sponsored search links

3
Keyword Auctions

5
Selling one Ad Slot $2 $5 $7 $3 Prospective advertisers

6
Selling one Ad Slot $2 $5 $7 $3 Pays $5 per click Vickrey Auction -Truthful - Efficient - Simple - …

7
Auction Model b1b1 b1b1 b2b2 b2b2 b3b3 b3b3 b4b4 b4b4 b5b5 b5b5 b6b6 b6b6 $$$ $$ $ $

8
Auction Model b1b1 b1b1 b2b2 b2b2 b3b3 b3b3 b4b4 b4b4 b5b5 b5b5 b6b6 b6b6 $$$ $$ $ $ Vickrey Auction VCG Auction Generalized Second Price Auction -Truthful - Efficient - Simple (?) - … - Not Truthful - Not Efficient - Even Simpler - …

9
Our Results Generalized Second Price Auction (GSP), although not optimal, has good social welfare guarantees: 1.618 for Pure Price of Anarchy 8 for Bayes-Nash Price of Anarchy GSP with uncertainty

10
(Simplified) Model α j : click-rate of slot j v i : value of player i b i : bid (declared value) Assumption: b i ≤ v i Since playing b i > v i is dominated strategy. α1α1 α2α2 α3α3 b1b1 b1b1 b2b2 b2b2 b3b3 b3b3 v1v1 v2v2 v3v3

11
(Simplified) Model v1v1 v2v2 v3v3 α1α1 α2α2 α3α3 b1b1 b1b1 b2b2 b2b2 b3b3 b3b3 pays b 1 per click

12
(Simplified) Model vivi αjαj bibi bibi j = σ(i)i = π(j) u i (b) = α σ(i) ( v i - b π(σ(i) + 1) ) Utility of player i : σ = π -1

13
Model vivi αjαj bibi bibi j = σ(i)i = π(j) σ = π -1 next highest bid u i (b) = α σ(i) ( v i - b π(σ(i) + 1) ) Utility of player i :

14
Model vivi αjαj bibi bibi j = σ(i)i = π(j) Nash equilibrium: σ = π -1 u i (b i,b -i ) ≥ u i (b’ i,b -i ) Is truth-telling always Nash ?

15
Example Non-truthful α 1 = 1 α 2 = 0.9 b 1 = 2 v 1 = 2 v 2 = 1 b 1 = 1 b 1 = 0.9 u 1 = 1 (2-1)u 1 = 0.9(2-0)

16
Measuring inefficiency vivi αjαj bibi bibi j = σ(i)i = π(j) σ = π -1 Social welfare = ∑ i v i α i ∑ i v i α σ(i) Optimal allocation =

17
Measuring inefficiency Price of Anarchy = max = Opt SW(Nash) Nash

18
Main Theorem 1 Thm: Pure Price of Anarchy ≤ 1.618 If b i ≤ v i and (b 1 …b n ) are bid in equilibrium, then for the allocation σ : ∑ i v i α σ(i) ≥ 1.618 -1 ∑ i v i α i Previously known [EOS, Varian]: Price of Stability = 1

19
GSP as a Bayesian Game Modeling uncertainty:

20
GSP as a Bayesian Game b b b ?

21
GSP as a Bayesian Game b b b b b b b b b b b b b b b b b b b b b b b b b b Idea: Optimize against a distribution.

22
Bayes-Nash solution concept Thm: Bayes-Nash PoA ≤ 8 Bayes-Nash models the uncertainty of other players about valuations Values v i are independent random vars Optimize against a distribution

23
Bayesian Model V1V1 V2V2 V3V3 v 1 ~ v 2 ~ v 3 ~ α1α1 α2α2 α3α3 b 1 (v 1 ) b 2 (v 2 ) b 3 (v 3 )

24
Model ViVi v i ~ αjαj b i (v i ) j = σ(i)i = π(j) E[u i (b i,b -i )|v i ] ≥ E[u i (b’ i,b -i )|v i ] Bayes-Nash equilibrium: Expectation over v -i

25
v i are random variables μ(i) = slot that player i occupies in Opt (also a random variable) Bayes-Nash PoA = Bayes-Nash Equilibrium E[ ∑ i v i α μ(i) ] E[ ∑ i v i α σ(i) ] Previously known [G-S]: Price of Stability ≠ 1

26
Sketch of the proof α2α2 α3α3 Opt α1α1 v1v1 v2v2 v3v3 αiαi α σ(j) vjvj v π(i) therefore: α σ(j) αiαi vjvj v π(i) ≥ 1 2 ≥ 1 2 or Simple and intuitive condition on matchings in equilibrium. α σ(j) αiαi vjvj v π(i) + ≥ 1

27
αiαi α σ(j) vjvj v π(i) α σ(j) αiαi vjvj v π(i) + ≥ 1 Need to show only for i π(j). It is a combination of 3 relations: α σ(j) ( v j – b π(σ(j)+1) ) ≥ α i ( v j – b π(i) ) [ Nash ] b π(σ(j)+1) ≥ 0 b π(i) ≤ v π(i) [conservative] Sketch of the proof

28
α σ(j) αiαi vjvj v π(i) + ≥ 1 2 SW = ∑ i α σ(i) v i + α i v π(i) = v π(i) = ∑ i α i v i ≥ α σ(i) αiαi vivi + ≥ ∑ i α i v i = Opt

29
Proof idea: new structural condition v i α σ(i) + α i v π(i) ≥ α i v i v i E[ α σ(i) |v i ] + E[ α μ( i ) v πμ (i) |v i ] ≥ ¼ v i E[ α μ( i ) |v i ] Bayes Nash version: Pure Nash version: αiαi α σ(j) vjvj v π(i)

30
Upcoming results [Lucier-Paes Leme-Tardos] Improved Bayes-Nash PoA to 3.164 Valid also for correlated distributions Future directions: Tight Pure PoA (we think it is 1.259)

Similar presentations

OK

6.853: Topics in Algorithmic Game Theory Fall 2011 Constantinos Daskalakis Lecture 21.

6.853: Topics in Algorithmic Game Theory Fall 2011 Constantinos Daskalakis Lecture 21.

© 2018 SlidePlayer.com Inc.

All rights reserved.

Ads by Google

Ppt on operating system deadlock Ppt on diversity in living organisms images Ppt on limits and continuity of functions Ppt on different types of occupations pictures Ppt on transportation in human beings higher Ppt on c programming algorithms and flowcharts Ppt on building information modeling degree Ppt on relations and functions for class 11th accountancy Ppt on mars one Ppt on centring prayer