Polyhedral Clinching Auctions and the AdWords Polytope Renato Paes - - PowerPoint PPT Presentation

▶
polyhedral clinching auctions
SMART_READER_LITE
LIVE PREVIEW

Polyhedral Clinching Auctions and the AdWords Polytope Renato Paes - - PowerPoint PPT Presentation

Aug 27, 2022 817 likes •1.45k views

Polyhedral Clinching Auctions and the AdWords Polytope Renato Paes Leme (Cornell University) Joint work with Gagan Goel and Vahab Mirrokni (Google NYC) Creating an Ads campaign Creating an Ads campaign Creating an Ads campaign How

slide-1
SLIDE 1

Polyhedral Clinching Auctions and the AdWords Polytope

Renato Paes Leme (Cornell University) Joint work with Gagan Goel and Vahab Mirrokni (Google NYC)

slide-2
SLIDE 2

Creating an Ads campaign …

slide-3
SLIDE 3

Creating an Ads campaign …

slide-4
SLIDE 4

Creating an Ads campaign …

slide-5
SLIDE 5

How to deal with budgets in practice ?

VCG, GSP, …

slide-6
SLIDE 6

How to deal with budgets in practice ?

VCG, GSP, … nice and well studied auction with good game-theoretic properties but without budgets…

slide-7
SLIDE 7

How to deal with budgets in practice ?

VCG, GSP, … budget layer

slide-8
SLIDE 8

How to deal with budgets in practice ?

VCG, GSP, … engineering fix to adapt the original auction to the budgeted setting. Original game theoretic analysis is now lost. budget layer

slide-9
SLIDE 9

How to deal with budgets in practice ?

VCG, GSP, … budget layer

slide-10
SLIDE 10

How to deal with budgets in practice ?

VCG, GSP, … control loop Polyhedral Clinching Auction Goal: Design an auction for AdWords that supports budgets natively, i.e., budgets are built in the game theoretic analysis

slide-11
SLIDE 11

What do we mean by budgets ?

slide-12
SLIDE 12

Classical quasi-linear utility function:

slide-13
SLIDE 13

Classical quasi-linear utility function: Budget constrained utility function:

slide-14
SLIDE 14

Classical quasi-linear utility function: Budget constrained utility function:

Very well understood: VCG, affine maximizers, …

slide-15
SLIDE 15

Classical quasi-linear utility function: Budget constrained utility function:

Very well understood: VCG, affine maximizers, … Surprisingly little is known.

slide-16
SLIDE 16

Goal: Design auctions for budget constrained agents

slide-17
SLIDE 17

Goal: Design auctions for budget constrained agents

  • Agents report values and budgets
  • Mechanism decides on allocation and

payments for each player

slide-18
SLIDE 18

Goal: Design auctions for budget constrained agents

  • Agents report values and budgets
  • Mechanism decides on allocation and

payments for each player

  • Requirements:
slide-19
SLIDE 19

Goal: Design auctions for budget constrained agents

  • Agents report values and budgets
  • Mechanism decides on allocation and

payments for each player

  • Requirements:

(feasible set)

slide-20
SLIDE 20

Desirable properties

  • Incentive Compatibility:

assumption: budgets Bi are public

  • Individual rationality:
  • Pareto optimality:

An outcome (x,p) is Pareto-optimal if there is no (x’,p’) such that u’i ≥ ui, Σp’i ≥ Σpi and at least one of them is strict.

slide-21
SLIDE 21

Our main contribution Solve this problem for a large class of feasible sets P.

slide-22
SLIDE 22

Our main contribution Solve this problem for a large class of feasible sets P : (scaled) polymatroids. Show this is impossible to be extended to general polytopes.

slide-23
SLIDE 23

Our main contribution Solve this problem for a large class of feasible sets P : (scaled) polymatroids. Show this is impossible to be extended to general polytopes. Conjecture: scaled polymatroids are the largest class for which this is possible. (we supply evidence for that)

slide-24
SLIDE 24

What do we know about budgets? [Dobzinski, Lavi, Nisan, FOCS’08] :: auction for one divisible good [Fiat, Leonardi, Saia, Sankowski, EC’11] :: auction for matching markets

slide-25
SLIDE 25

What do we know about budgets? [Dobzinski, Lavi, Nisan, FOCS’08] :: auction for one divisible good [Fiat, Leonardi, Saia, Sankowski, EC’11] :: auction for matching markets based on the clinching auctions framework [Ausubel, AER’97]

slide-26
SLIDE 26

How does it fit in our goal ? [Dobzinski, Lavi, Nisan, FOCS’08] P = Uniform Matroid [Fiat, Leonardi, Saia, Sankowski, EC’11] P = Transversal Matroid

slide-27
SLIDE 27

How does it fit in our goal ? [Dobzinski, Lavi, Nisan, FOCS’08] P = Uniform Matroid [Fiat, Leonardi, Saia, Sankowski, EC’11] P = Transversal Matroid For AdWords and other more complicated markets, we need to solve it for more generic feasibility constraints P

slide-28
SLIDE 28

Our Results We provide an auction with all the desirable properties for any polymatroid P.

slide-29
SLIDE 29

Our Results We provide an auction with all the desirable properties for any polymatroid P.

  • Incentive compatibility
  • Individual Rationality
  • Budget Feasibility
  • Pareto Optimality
slide-30
SLIDE 30

Our Results We provide an auction with all the desirable properties for any polymatroid P. for a submodular function f.

slide-31
SLIDE 31

Our Results We provide an auction with all the desirable properties for any polymatroid P. Our auction only needs oracle access to the submodular function f. Our auction has a natural geometric flavor.

slide-32
SLIDE 32

Our Results We provide an auction with all the desirable properties for any polymatroid P. Many applications Auctions for network design, queuing systems, video on demand, matching markets, internet advertisement, …

slide-33
SLIDE 33

Our Results We provide an auction with all the desirable properties for any polymatroid P. Many applications Auctions for network design, queuing systems, video on demand, matching markets, internet advertisement, …

slide-34
SLIDE 34

Our results The set of that can be

  • btained this way form a polymatroid. We

call it the AdWords Polytope. General model:

  • multiple slots
  • multiple keywords
  • easy to generalize
slide-35
SLIDE 35

Also on Sponsored Search with Budgets Independently, [Colini-Baldeschi, Henzinger, Leonardi, Starnberger, 2012] design an auction for sponsored search with one keyword, multiple slots and budgets.

slide-36
SLIDE 36

Our auction

polytope of feasible allocations price clock

slide-37
SLIDE 37

Our auction

slide-38
SLIDE 38

Our auction

slide-39
SLIDE 39

Our auction

slide-40
SLIDE 40

Our auction

slide-41
SLIDE 41

Our auction In each step compute demands at price if ; and o.w.

slide-42
SLIDE 42

Our auction In each step compute demands at price Compute clinched amount if ; and o.w.

slide-43
SLIDE 43

Computing clinched amounts What is the allocations that are still feasible at this point?

slide-44
SLIDE 44

Our auction: how to implement clinch ? How much can I allocate to 1 without harming player 2?

slide-45
SLIDE 45

Our auction: how to implement clinch ?

slide-46
SLIDE 46

Our auction: how to implement clinch ?

Clinching step

slide-47
SLIDE 47

Our auction: how to implement clinch ?

Clinching step Theorem: Clinching as defined above results in a feasible

  • allocation. If P is a polymatroid, δi can be computed efficiently

using submodular minimization.

slide-48
SLIDE 48

Our auction: how to implement clinch ?

Clinching step Theorem: Clinching as defined above results in a feasible

  • allocation. If P is a polymatroid, δi can be computed efficiently

using submodular minimization. [in practice there are more efficient algorithms for each case]

slide-49
SLIDE 49

Summary of the proof

  • Show clinching is well-defined and can

be computed efficiently

  • Characterize Pareto-optimal outcomes

for polymatroidal environments

  • Show that the auction produces an
  • utcome satisfying the characterization
slide-50
SLIDE 50

Extensions and Limits Going beyond polymatroids…

slide-51
SLIDE 51

General convex environment One budget-constrained player For a single budget constrained player (and many other unconstrained ones), it is possible do design an auction for any convex environment.

slide-52
SLIDE 52

What about 2 budget constrained players ? Weak impossibility: There is no auction following the clinching framework beyond (scaled) polymatroids.

slide-53
SLIDE 53

What about 2 budget constrained players ? Weak impossibility: There is no auction following the clinching framework beyond (scaled) polymatroids. Stronger impossibility: There exists a class

  • f polytopes, for which no auction exists

satisfying all the desirable properties.

slide-54
SLIDE 54

What about 2 budget constrained players ? Weak impossibility: There is no auction following the clinching framework beyond (scaled) polymatroids. Stronger impossibility: There exists a class

  • f polytopes, for which no auction exists

satisfying all the desirable properties.

No hope of an auction for a general polyhedral environment.

slide-55
SLIDE 55

Impossibility for decreasing marginals Single divisible good: Decreasing marginal valuations

slide-56
SLIDE 56

Impossibility for decreasing marginals Single divisible good: Thm: No auction with all the desirable properties for one divisible good with decreasing marginals.

Strengthens previous impossibility results of [Lavi, May’11] and [Fiat et al’11]

slide-57
SLIDE 57

Summary

Clinching auction for polymatroids

slide-58
SLIDE 58

Summary

Clinching auction for polymatroids Characterization of Pareto Optimal Auctions in general polyhedral environments

slide-59
SLIDE 59

Summary

Clinching auction for polymatroids Impossibility for decreasing-marginals and budgets Characterization of Pareto Optimal Auctions in general polyhedral environments Impossibility for general polytopes

slide-60
SLIDE 60

Summary

Clinching auction for polymatroids Impossibility for decreasing-marginals and budgets Characterization of Pareto Optimal Auctions in general polyhedral environments Impossibility for general polytopes Positive results for one budget-constr agent and general environments

slide-61
SLIDE 61

Thanks !