some extensions to mechanism design -...
TRANSCRIPT
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Some Extensions to Mechanism Design
Sujit Prakash Gujar
Thesis Supervisor : Y Narahari
Department of Computer Science and AutomationIndian Institute of Science
Bangalore-12
January 19, 2008
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 1 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Thesis Overview
Extending mechanism design to general situations motivated by realworld problems
Mechanism design with heterogeneous objects (multi dimensionalprivate information) is formidable challenge
Reducing budget imbalance1
Optimal multi unit combinatorial auction
1H. Moulin. “Efficient, strategy-proof and almost budget-balanced assignment”.Technical report, 2007.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 2 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Publications Based On Work
“Almost Budget Balanced Mechanism Design”, Sujit Gujar and YNarahari, Working Paper.
“Foundations of Mechanism Design: A Tutorial, Part 1: KeyConcepts and Classical Results”, Dinesh Garg, Y Narahari and SujitGujar. Sadhana - Indian Academy Proceedings in EngineeringSciences, to appear 2008.
“Foundations of Mechanism Design: A Tutorial, Part 2: AdvancedConcepts and Results”, Dinesh Garg, Y Narahari and Sujit Gujar.Sadhana - Indian Academy Proceedings in Engineering Sciences, toappear 2008.
“An Optimal Multi-Unit Combinatorial Procurement Auction withSingle Minded Bidders”, Sujit Gujar and Y Narahari, Submitted toManaging Complexity in Distributed World, MCDES 2008.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 3 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Publications Based On Work
“Almost Budget Balanced Mechanism Design”, Sujit Gujar and YNarahari, Working Paper.
“Foundations of Mechanism Design: A Tutorial, Part 1: KeyConcepts and Classical Results”, Dinesh Garg, Y Narahari and SujitGujar. Sadhana - Indian Academy Proceedings in EngineeringSciences, to appear 2008.
“Foundations of Mechanism Design: A Tutorial, Part 2: AdvancedConcepts and Results”, Dinesh Garg, Y Narahari and Sujit Gujar.Sadhana - Indian Academy Proceedings in Engineering Sciences, toappear 2008.
“An Optimal Multi-Unit Combinatorial Procurement Auction withSingle Minded Bidders”, Sujit Gujar and Y Narahari, Submitted toManaging Complexity in Distributed World, MCDES 2008.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 3 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Publications Based On Work
“Almost Budget Balanced Mechanism Design”, Sujit Gujar and YNarahari, Working Paper.
“Foundations of Mechanism Design: A Tutorial, Part 1: KeyConcepts and Classical Results”, Dinesh Garg, Y Narahari and SujitGujar. Sadhana - Indian Academy Proceedings in EngineeringSciences, to appear 2008.
“Foundations of Mechanism Design: A Tutorial, Part 2: AdvancedConcepts and Results”, Dinesh Garg, Y Narahari and Sujit Gujar.Sadhana - Indian Academy Proceedings in Engineering Sciences, toappear 2008.
“An Optimal Multi-Unit Combinatorial Procurement Auction withSingle Minded Bidders”, Sujit Gujar and Y Narahari, Submitted toManaging Complexity in Distributed World, MCDES 2008.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 3 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Agenda
1 Introduction to Mechanism Design and AuctionsMechanism DesignAuctions
2 State of the ArtMyerson’s WorkOptimal Auctions Beyond Myerson
3 An Optimal Multi Unit Combinatorial AuctionAssumptionsNotationNecessary and Sufficient Conditions for Bayesian IncentiveCompatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
4 Conclusion and Future Work
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 4 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Mechanism DesignAuctions
Agenda
1 Introduction to Mechanism Design and AuctionsMechanism DesignAuctions
2 State of the ArtMyerson’s WorkOptimal Auctions Beyond Myerson
3 An Optimal Multi Unit Combinatorial AuctionAssumptionsNotationNecessary and Sufficient Conditions for Bayesian IncentiveCompatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
4 Conclusion and Future Work
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 5 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Mechanism DesignAuctions
Mechanism Design
Mechanism Design is the art of designing rules of a game to achievea specific outcome in presence of multiple self-interested agents,each with private information about their preferences.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 6 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Mechanism DesignAuctions
Mechanism Design
Mechanism Design is the art of designing rules of a game to achievea specific outcome in presence of multiple self-interested agents,each with private information about their preferences.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 6 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Mechanism DesignAuctions
Agenda
1 Introduction to Mechanism Design and AuctionsMechanism DesignAuctions
2 State of the ArtMyerson’s WorkOptimal Auctions Beyond Myerson
3 An Optimal Multi Unit Combinatorial AuctionAssumptionsNotationNecessary and Sufficient Conditions for Bayesian IncentiveCompatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
4 Conclusion and Future Work
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 7 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Mechanism DesignAuctions
Auctions
First Price Auction (FPA) for selling a single item.
The bidder with the highest bid wins.She pays as much as she has bid for.
Second Price Auction (SPA) for selling a single item.
The bidder with the highest bid wins.She pays as much as the second highest bid.
Vickrey 2 showed : The truth revelation is dominant strategy insecond price auction.
2W. Vickrey. Counter speculation, auctions, and competitive sealed tenders.Journal of Finance, 16(1):8-37, March 1961.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 8 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Mechanism DesignAuctions
Auctions
First Price Auction (FPA) for selling a single item.
The bidder with the highest bid wins.
She pays as much as she has bid for.
Second Price Auction (SPA) for selling a single item.
The bidder with the highest bid wins.She pays as much as the second highest bid.
Vickrey 2 showed : The truth revelation is dominant strategy insecond price auction.
2W. Vickrey. Counter speculation, auctions, and competitive sealed tenders.Journal of Finance, 16(1):8-37, March 1961.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 8 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Mechanism DesignAuctions
Auctions
First Price Auction (FPA) for selling a single item.
The bidder with the highest bid wins.She pays as much as she has bid for.
Second Price Auction (SPA) for selling a single item.
The bidder with the highest bid wins.She pays as much as the second highest bid.
Vickrey 2 showed : The truth revelation is dominant strategy insecond price auction.
2W. Vickrey. Counter speculation, auctions, and competitive sealed tenders.Journal of Finance, 16(1):8-37, March 1961.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 8 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Mechanism DesignAuctions
Auctions
First Price Auction (FPA) for selling a single item.
The bidder with the highest bid wins.She pays as much as she has bid for.
Second Price Auction (SPA) for selling a single item.
The bidder with the highest bid wins.She pays as much as the second highest bid.
Vickrey 2 showed : The truth revelation is dominant strategy insecond price auction.
2W. Vickrey. Counter speculation, auctions, and competitive sealed tenders.Journal of Finance, 16(1):8-37, March 1961.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 8 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Mechanism DesignAuctions
Auctions
First Price Auction (FPA) for selling a single item.
The bidder with the highest bid wins.She pays as much as she has bid for.
Second Price Auction (SPA) for selling a single item.
The bidder with the highest bid wins.
She pays as much as the second highest bid.
Vickrey 2 showed : The truth revelation is dominant strategy insecond price auction.
2W. Vickrey. Counter speculation, auctions, and competitive sealed tenders.Journal of Finance, 16(1):8-37, March 1961.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 8 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Mechanism DesignAuctions
Auctions
First Price Auction (FPA) for selling a single item.
The bidder with the highest bid wins.She pays as much as she has bid for.
Second Price Auction (SPA) for selling a single item.
The bidder with the highest bid wins.She pays as much as the second highest bid.
Vickrey 2 showed : The truth revelation is dominant strategy insecond price auction.
2W. Vickrey. Counter speculation, auctions, and competitive sealed tenders.Journal of Finance, 16(1):8-37, March 1961.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 8 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Mechanism DesignAuctions
Auctions
First Price Auction (FPA) for selling a single item.
The bidder with the highest bid wins.She pays as much as she has bid for.
Second Price Auction (SPA) for selling a single item.
The bidder with the highest bid wins.She pays as much as the second highest bid.
Vickrey 2 showed : The truth revelation is dominant strategy insecond price auction.
2W. Vickrey. Counter speculation, auctions, and competitive sealed tenders.Journal of Finance, 16(1):8-37, March 1961.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 8 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Mechanism DesignAuctions
Space of Mechanisms
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 9 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Myerson’s WorkOptimal Auctions Beyond Myerson
Agenda
1 Introduction to Mechanism Design and AuctionsMechanism DesignAuctions
2 State of the ArtMyerson’s WorkOptimal Auctions Beyond Myerson
3 An Optimal Multi Unit Combinatorial AuctionAssumptionsNotationNecessary and Sufficient Conditions for Bayesian IncentiveCompatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
4 Conclusion and Future Work
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 10 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Myerson’s WorkOptimal Auctions Beyond Myerson
Optimal Auction
Myerson3: Introduced the notion of “Optimal auction”
Maximizes revenue to the sellerSatisfies interim individual rationalityBayesian incentive compatibility
3R. B. Myerson. Optimal auction design. Mathematics of Operations Research,6(1):58-73, February 1981
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 11 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Myerson’s WorkOptimal Auctions Beyond Myerson
Optimal Auction
Myerson3: Introduced the notion of “Optimal auction”
Maximizes revenue to the seller
Satisfies interim individual rationalityBayesian incentive compatibility
3R. B. Myerson. Optimal auction design. Mathematics of Operations Research,6(1):58-73, February 1981
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 11 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Myerson’s WorkOptimal Auctions Beyond Myerson
Optimal Auction
Myerson3: Introduced the notion of “Optimal auction”
Maximizes revenue to the sellerSatisfies interim individual rationality
Bayesian incentive compatibility
3R. B. Myerson. Optimal auction design. Mathematics of Operations Research,6(1):58-73, February 1981
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 11 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Myerson’s WorkOptimal Auctions Beyond Myerson
Optimal Auction
Myerson3: Introduced the notion of “Optimal auction”
Maximizes revenue to the sellerSatisfies interim individual rationalityBayesian incentive compatibility
3R. B. Myerson. Optimal auction design. Mathematics of Operations Research,6(1):58-73, February 1981
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 11 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Myerson’s WorkOptimal Auctions Beyond Myerson
Optimal Auction
Myerson3: Introduced the notion of “Optimal auction”
Maximizes revenue to the sellerSatisfies interim individual rationalityBayesian incentive compatibility
3R. B. Myerson. Optimal auction design. Mathematics of Operations Research,6(1):58-73, February 1981
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 11 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Myerson’s WorkOptimal Auctions Beyond Myerson
Agenda
1 Introduction to Mechanism Design and AuctionsMechanism DesignAuctions
2 State of the ArtMyerson’s WorkOptimal Auctions Beyond Myerson
3 An Optimal Multi Unit Combinatorial AuctionAssumptionsNotationNecessary and Sufficient Conditions for Bayesian IncentiveCompatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
4 Conclusion and Future Work
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 12 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Myerson’s WorkOptimal Auctions Beyond Myerson
Optimal Auctions Beyond Myerson
Armstrong, M. Optimal multi-object auctions. Review of EconomicStudies 67, 3 (July 2000), 455-81.
Kumar, A., and Iyengar, G. Optimal procurement auctions fordivisible goods with capacitated suppliers. Tech. rep., ColumbiaUniversity, 2006. Technical Report TR-2006-01.
Gautam, R. K., Hemachandra, N., Narahari, Y., and Prakash, H.Optimal auctions for multi-unit procurement with volume discountbids. Proceedings of IEEE Conference on E-Commerce Technology,CEC-2007, Tokyo, Japan (2007), 21-28.
Ledyard, J. O. Optimal combinatoric auctions with single-mindedbidders. In EC’07: Proceedings of the 8th ACM conference onElectronic commerce (New York, NY, USA, 2007), ACM Press, pp.237-242.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 13 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
Optimal Multi Unit CombinatorialAuction in the Presence of Single
Minded, Capacitated Bidders
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 14 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
Agenda
1 Introduction to Mechanism Design and AuctionsMechanism DesignAuctions
2 State of the ArtMyerson’s WorkOptimal Auctions Beyond Myerson
3 An Optimal Multi Unit Combinatorial AuctionAssumptionsNotationNecessary and Sufficient Conditions for Bayesian IncentiveCompatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
4 Conclusion and Future Work
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 15 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
Assumptions
1 The sellers are single minded
2 The sellers can collectively fulfill the demands specified by the buyer
3 The sellers are capacitated
4 The seller will never inflate his capacity(This is an important assumption)
5 All the participants are rational and intelligent
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 16 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
Agenda
1 Introduction to Mechanism Design and AuctionsMechanism DesignAuctions
2 State of the ArtMyerson’s WorkOptimal Auctions Beyond Myerson
3 An Optimal Multi Unit Combinatorial AuctionAssumptionsNotationNecessary and Sufficient Conditions for Bayesian IncentiveCompatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
4 Conclusion and Future Work
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 17 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
Notation
I Set of items the buyer is interested in buying, {1, 2, . . . , m}Dj Demand for item j , j = . . . m
N Set of sellers. {1, 2, . . . , n}ci True cost of production of one unit of bundle of interest to the seller i ,
ci ∈ [ci , ci ]
qi True capacity for bundle which seller i can supply, qi ∈ [qi , qi ]
ci Reported cost by the seller i
qi Reported capacity by the seller i
bi Bid of the seller i . bi = (ci , qi )
b Bid vector, (b1, b2, . . . , bn)
b−i Bid vector without the seller i , i.e. (b1, b2, . . . , bi−1, bi+1, . . . , bn)
ti (b) Payment to the seller i when submitted bid vector is b
Ti (bi ) Expected payment to the seller i when he submits bid bi .Expectation is taken over all possible values of b−i
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 18 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
Notation
xi = xi (b) Quantity of the bundle to be procured from the seller iwhen the bid vector is b
Xi (bi ) Expected quantity of the bundle to be procured from the seller iwhen he submits bid bi .Expectation is taken over all possible values of b−i
fi (ci , qi ) Joint probability density function of (ci , qi )
Fi (ci , qi ) Cumulative distribution function of fi (ci , qi )
fi (ci |qi ) Conditional probability density function of production costwhen it is given that the capacity of the seller i is qi
Fi (ci |qi ) Cumulative distribution function of fi (ci |qi )
Hi (ci , qi ) Virtual cost function for seller i ,
Hi (ci , qi ) = ci + Fi (ci |qi )fi (ci |qi )
ρi (bi ) Expected offered surplus to seller i , when his bid is bi
Table: Notation
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 19 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
Agenda
1 Introduction to Mechanism Design and AuctionsMechanism DesignAuctions
2 State of the ArtMyerson’s WorkOptimal Auctions Beyond Myerson
3 An Optimal Multi Unit Combinatorial AuctionAssumptionsNotationNecessary and Sufficient Conditions for Bayesian IncentiveCompatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
4 Conclusion and Future Work
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 20 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
Incentive Compatibility
× Sellers may not be willing to reveal their true types.
√Offer them incentives for reporting true costs and capacities.
We propose the following incentive structure, ∀i ∈ N,
ρi (bi ) = Ti (bi )− ciXi (bi ), where bi = (ci , qi )
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 21 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
Incentive Compatibility
× Sellers may not be willing to reveal their true types.√Offer them incentives for reporting true costs and capacities.
We propose the following incentive structure, ∀i ∈ N,
ρi (bi ) = Ti (bi )− ciXi (bi ), where bi = (ci , qi )
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 21 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
Incentive Compatibility
× Sellers may not be willing to reveal their true types.√Offer them incentives for reporting true costs and capacities.
We propose the following incentive structure, ∀i ∈ N,
ρi (bi ) = Ti (bi )− ciXi (bi ), where bi = (ci , qi )
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 21 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
Necessary and Sufficient Conditions for Bayesian IncentiveCompatibility and Individual Rationality
We proved,
Theorem 1
Any mechanism in the presence of single minded, capacitated sellers isBIC and IR iff
1 ρi (bi ) = ρi (ci ,qi ) +∫ ci
ciXi (t, qi )dt
2 ρi (bi ) non-negative, and non-decreasing in qi ∀ ci ∈ [ci , ci ]
3 The quantity which seller i is asked to supply, Xi (ci , qi ) isnon-increasing in ci ∀qi ∈ [qi , qi ].
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 22 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
Agenda
1 Introduction to Mechanism Design and AuctionsMechanism DesignAuctions
2 State of the ArtMyerson’s WorkOptimal Auctions Beyond Myerson
3 An Optimal Multi Unit Combinatorial AuctionAssumptionsNotationNecessary and Sufficient Conditions for Bayesian IncentiveCompatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
4 Conclusion and Future Work
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 23 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
An Optimal Auction
The buyer’s problem is to solve,
min Eb
∑ni=1 ti (b) s.t.
1 ti (b) = ρi (b) + cixi (b)
2 All three conditions in Theorem 1 hold true.
3 She procures at least Dj units of each item j .
We have shown that an optimal auction for the buyer in the presence ofthe single minded sellers is,
Optimal Auction
min∫ q
q
∫ c
c
(∑ni=1 Hi (ci , qi )xi (ci , qi )
)f (c , q)dc dq s.t.
1. ∀; i , Xi (ci , qi ) is non-increasing in ci ,∀ qi .2. The Buyer’s minimum requirement of each item is satisfied.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 24 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
An Optimal Auction
The buyer’s problem is to solve,
min Eb
∑ni=1 ti (b) s.t.
1 ti (b) = ρi (b) + cixi (b)
2 All three conditions in Theorem 1 hold true.
3 She procures at least Dj units of each item j .
We have shown that an optimal auction for the buyer in the presence ofthe single minded sellers is,
Optimal Auction
min∫ q
q
∫ c
c
(∑ni=1 Hi (ci , qi )xi (ci , qi )
)f (c , q)dc dq s.t.
1. ∀; i , Xi (ci , qi ) is non-increasing in ci ,∀ qi .2. The Buyer’s minimum requirement of each item is satisfied.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 24 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
Agenda
1 Introduction to Mechanism Design and AuctionsMechanism DesignAuctions
2 State of the ArtMyerson’s WorkOptimal Auctions Beyond Myerson
3 An Optimal Multi Unit Combinatorial AuctionAssumptionsNotationNecessary and Sufficient Conditions for Bayesian IncentiveCompatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
4 Conclusion and Future Work
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 25 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
Regularity Assumption
Regularity Assumption
Hi (ci , qi ) = ci +Fi (ci |qi )
fi (ci |qi )
is non-increasing in qi and non-decreasing in ci .
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 26 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
AssumptionsNotationNecessary and Sufficient Conditions for Bayesian Incentive Compatibility and Individual RationalityAn Optimal AuctionAn Optimal Auction : Under Regularity Assumption
An Optimal auction : Under regularity Assumption
The buyer’s optimal auction is,
minn∑
i=1
xiHi (ci , qi ) subject to
1 0 ≤ xi ≤ qi
2 Buyer’s demands are satisfied.
The buyer pays each seller i the amount
ti = cix∗i +
∫ ci
ci
xi (t, qi )dt (1)
where x∗i is what agent i has to supply after solving the above problem.Note: This auction enjoys Dominant Strategy Incentive Compatibility.
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 27 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Summary So Far...
We have seen,
Necessary and sufficient conditions for BIC and individual rationality
Characterization of an optimal multi unit combinatorial procurementauction in the presence of single minded capacitated bidders
An optimal auction, for the same, which is dominant strategyincentive compatible if some regularity condition holds true
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 28 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Directions for Future Work
Design of an optimal combinatorial auction for more general settingis a formidable challenge
Extending Moulin’s work for heterogeneous objects
Repeated/Stochastic mechanism design
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 29 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Questions?
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 30 / 31
Introduction to Mechanism Design and AuctionsState of the Art
An Optimal Multi Unit Combinatorial AuctionConclusion and Future Work
Thank You!!!
Sujit Prakash Gujar (CSA, IISc) Some Extensions to Mechanism Design January 19, 2008 31 / 31