Parkes, D. 1999b. Optimal auction design for agents with hard valuation problems. In Agent-Mediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....Can the bidders use this information to regulate their information about their own values Probably, but how much do they win If one cannot guarantee that deliberate loss of privacy will decrease the cognitive costs, it is better not to lose any. Cognitive cost is modeled in some publications [Par99,LS01], but there the authors are more concerned about the agents doing the computations, not the human beings. Therefore, we argue that when constructing an online auction mechanism, one should first make sure that the auction is Pareto efficient, correct and (almost ideally) privacy preserving. The ....

....cognitive costs in online auctions is by Parkes, Ungar and Foster [PUF98] Their paper analyzed the existing mechanisms from this aspect and concluded that the English auctions are the best in the context of bounded rationality. It has followed by a large body of research in this direction, see [Par99,LS01] for some examples and further references. However, most of the papers in this line of research do not actually propose new mechanisms, but instead propose criteria on how to choose between already existing and well known mechanisms. Moreover, the mentioned papers are more concerns about fully ....

[Article contains additional citation context not shown here]

David C. Parkes. Optimal Auction Design for Agents with Hard Valuation Problems. In Alexandros Moukas, Carles Sierra, and Fredrik Ygge, editors, Agent Mediated Electronic Commerce II, Towards Next-Generation Agent-Based Electronic Commerce Systems, IJCAI 1999 Workshop, volume 1788 of Lecture Notes in Computer Science, pages 206--219. Springer-Verlag, 1999.

....received much less attention, is that of bidding. There are 2 1 bundles, and each bidder may need to bid on all of them to fully express its preferences. This can be undesirable for any of several reasons: 1a) determining one s valuation for any given bundle can be computationally intractable [21, 23, 17, 14]; 1b) there is a huge number of bundles to evaluate; 2) communicating the bids can incur prohibitive overhead (e.g. network traffic) and (3) bidders may prefer Dr. Sandholm s work was funded by, and conducted at, CombineNet, Inc. 311 S. Craig St. Pittsburgh, PA 15213. not to reveal all of ....

David C Parkes. Optimal auction design for agents with hard valuation problems. In AgentMediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence, Stockholm, Sweden, 1999.

....has received much less attention, is that of bidding. There are 2 1 bundles, and each agent may need to bid on all of them to fully express its preferences. This can be undesirable for any of several reasons: determining one s valuation for any given bundle can be computationally intractable [9, 13,17]; there is a huge number of bundles to evaluate; communicating the bids can incur prohibitive overhead (e.g. network traffic) and agents may prefer not to reveal all of their valuation information due to reasons of privacy or long term competitiveness [16] Appropriate bidding languages [7, 8, ....

D. C. Parkes. Optimal auction design for agents with hard valuation problems. AgentMediated Electronic Commerce Workshop at IJCAI, 1999.

....has received much less attention, is that of bidding. There are # bundles, and each agent may need to bid on all of them to fully express its preferences. This can be undesirable for any of several reasons: determining one s valuation for any given bundle can be computationally intractable [10,11,16,21, 23]; there is a huge number of bundles to evaluate; communicating the bids can incur prohibitive overhead (e.g. network traffic) and agents may prefer not to reveal all of their valuation information due to reasons of privacy or long term competitiveness [20] Appropriate bidding languages [7, 9, ....

David C Parkes. Optimal auction design for agents with hard valuation problems. In Agent-Mediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence, Stockholm, Sweden, 1999.

....received much less attention, is that of bidding. There are 2 1 bundles, and each agent may need to bid on all of them to fully express its preferences. This can be undesirable for any of several reasons: 1a) determining one s valuation for any given bundle can be computationally intractable [19, 21, 15, 12]; 1b) there is a huge number of bundles to evaluate; 2) communicating the bids can incur prohibitive overhead (e.g. network traffic) and (3) agents may prefer not to reveal all of their valuation information due to reasons of privacy or long term competitiveness. Appropriate bidding languages ....

David C Parkes. Optimal auction design for agents with hard valuation problems. In AgentMediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence, Stockholm, Sweden, 1999.

....of mechanisms that lead to as efficient as possible outcomes with as little redundant computation as possible. While it has been pointed out that the central revelation principle from noncooperative game theory [19] ceases to apply when computational complexity limits each agent s rationality [23,32], our model provides a framework for actually deriving results in settings where the revelation principle fails to hold. This framework allows one to analyze problems beyond bargaining as well, including auctions where each agent needs to potentially solve an intractable problem to determine its ....

D.C. Parkes, Optimal auction design for agents with hard valuation problems, in: Proc. Agent-Mediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence, Stockholm, Sweden, 1999.

....has received much less attention, is that of bidding. There are 2 1 bundles, and each agent may need to bid on all of them to fully express its preferences. This can be undesirable for any of several reasons: determining one s valuation for any given bundle can be computationally intractable [10, 11,16,21, 23]; there is a huge number of bundles to evaluate; communicating the bids can incur prohibitive overhead (e.g. network traffic) and agents may prefer not to reveal all of their valuation information due to reasons of privacy or long term competitiveness [20] Appropriate bidding languages [7, 9, ....

David C Parkes. Optimal auction design for agents with hard valuation problems. In Agent-Mediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence, Stockholm, Sweden, 1999.

....such as shortest paths and scheduling on unrelated machines, and raised some associated computational issues. The papers [11] and [18] address computational di#culties with the VCG mechanism. The topic of revenue maximizing auctions has received a huge amount of attention. See the recent papers [1, 3, 6, 7, 8, 14, 16, 21, 22], or [5] for a recent survey on combinatorial auctions. The issue of frugality is raised in [2] which presents a truthful mechanism for minimizing makespan on machines with speeds. Their mechanism gives a 3 approximation and pays the machines only a logarithmic factor more than the actual costs ....

David C. Parkes. Optimal auction design for agents with hard valuation problems. In Alexandros Moukas, Carles Sierra, and Fredrik Ygge, editors, Agent Mediated Electronic Commerce II, Towards Next-Generation Agent-Based Electronic Commerce Systems, IJCAI

....Sandholm noted that under a model of costly computation, the dominant strategy property of Vickrey auctions fails to hold [21] Instead, an agent s best deliberation action can depend on the other agents. In recent work, auction settings where agents have hard valuation problems have been studied [16]. Auction design is presented as a way to simplify the meta deliberation problems of the agents , with the goal of providing incentives for the right agents to deliberate for the right amount of time. A costly computation model, with simple deliberation control where agents compute to refine ....

David C Parkes. Optimal auction design for agents with hard valuation problems. In Agent-Mediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence, Stockholm, 1999.

....of mechanisms that lead to as efficient as possible outcomes with as little redundant computation as possible. While it has been pointed out that the central revelation principle from noncooperative game theory [19] ceases to apply when computational complexity limits each agent s rationality [32, 23], our model provides a framework for actually deriving results in settings where the revelation principle fails to hold. This framework allows one to analyze problems beyond bargaining as well, including auctions where each agent needs to potentially solve an intractable problem to determine its ....

David C Parkes. Optimal auction design for agents with hard valuation problems. In Agent-Mediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence, Stockholm, Sweden, 1999.

....that under a model of costly computation, the dominant strategy property of Vickrey auctions fails to hold [Sandholm, 2000] Instead, an agent s best deliberation action can depend on the other agents. In recentwork auction settings where agents have hard valuation problems have been studied [Parkes, 1999] . Auction design is presented as a way to simplify the meta deliberation problems of the agents , providing incentives for the right agents to deliberate for the right amountof time. A costly computation model where agents compute to re ne their valuations is used. However, situations where ....

David C Parkes. Optimal auction design for agents with hard valuation problems. In Agent-Mediated Electronic Commerce Workshop at the International Joint ConferenceonArticial Intelligence, Stockholm, Sweden, 1999.

....Another problem, which has received less attention, is that combinatorial auctions require potentially every bundle to be bid on, and there are exponentially many bundles. This is complex for the bidder because she may need to invest effort or computation into determining each of her valuations [12, 4, 5, 7]. It can also be undesirable from the perspective of revealing unnecessary private information and from the perspective of unnecessary communication. The key observation of this paper is that topological structure that is inherent in the problem can be used to intelligently ask only relevant ....

....it leads to an incentive compatible mechanism when used in conjunction with our preference elicitation method. Another interesting avenue is to integrate this selective revelation technique with open cry ascending combinatorial auctions, where some unnecessary revelation is avoided in another way [9, 8, 6, 18, 7]. Namely, if the price of a bundle is already too high for an agent for sure, the agent need not compute or communicate the exact value. Combining that idea with the topological ideas of this paper is likely to lead to a hybrid protocol that requires less information than either method alone. ....

David C Parkes. Optimal auction design for agents with hard valuation problems. In Agent-Mediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence, Stockholm, Sweden, 1999.

....Another problem, which has received less attention, is that combinatorial auctions require potentially every bundle to be bid on, and there are exponentially many bundles. This is complex for the bidder because she may need to invest e#ort or computation into determining each of her valuations [12, 14, 5, 6, 8]. It can also be undesirable from the perspective of revealing unnecessary private information and from the perspective of unnecessary communication. # This work was funded by, and conducted at, CombineNet, Inc. 311 S. Craig St. Pittsburgh, PA 15213. Permission to make digital or hard copies ....

....have (about the others valuations) so in some cases the query directed revelation is e#ective as shown in the paragraph before last. We can also integrate the elicitation technique with opencry ascending combinatorial auctions, where some unnecessary revelation is avoided via price feedback [10, 9, 7, 18, 8]. Namely, if the price of a bundle is already too high for an agent, the agent need not compute or communicate her exact valuation. On top of that, the elicitor can guide revelation, and bidders can answer queries that were not asked. 5. INCENTIVECOMPATIBLEELICITATION Motivating the bidders to ....

David Parkes. Optimal auction design for agents with hard valuation problems. In Agent-Mediated Electronic Commerce Workshop at the IJCAI, 1999.

No context found.

D. C. Parkes. Optimal auction design for agents with hard valuation problems. In Proc. IJCAI-99 Workshop on Agent Mediated Electronic Commerce, pages 206--219, July 1999. Stockholm.

No context found.

David C Parkes. Optimal auction design for agents with hard valuation problems. In Proc. IJCAI-99 Workshop on Agent Mediated Electronic Commerce, July 1999. Stockholm.

....study of a POMDP model to construct offline approximations to sequentially optimal preference elicitation policies, again in single agent environments. The role of iterative auctions in addressing preference elicitation costs has been previously modeled for single item and multi item problems [Par99, Par01]. Recent theoretical analysis [CJ00] for a simple preference elicitation model in which agents can refine their valuations once, also supports the allocative efficiency benefits of iterative auctions for agents with costly preference elicitation. A key motivation for ascending price ....

David C Parkes. Optimal auction design for agents with hard valuation problems. In Proc. IJCAI-99 Workshop on Agent Mediated Electronic Commerce, pages 206--219, July 1999. Stockholm.

....has practical significance because it addresses the computational and informational complexity of bundle auctions. For agents, the auction can terminate before agents have revealed (or even computed) their values for every bundle. This can be important when agents have hard valuation problems [15, 16]. Furthermore, the auc tioneer only needs to generate explicit prices on a subset of bundles and often solves smaller winner determination problems than in the classic sealed bid bundle auction (the Generalized Vickrey Auction [23] Also relevant, because the winner determination problem remains ....

....to tradeoff allocative efficiency for computation and communication cost will be important in practical applications. Iterative auctions can allow more efficient computation than sealed bid auctions when agents have hard valuation problems, and this can lead to higher allocative efficiency [15]. Bundle can generate solutions without information from agents on their value for every bundle. Agents only need to determine the bundles that maximize utility given prices to be able follow a best response bidding strategy, and this can be done with approximate values for bundles. Although ....

Parkes, D.C. Optimal auction design for agents with hard valuation problems. In Proc. 2nd Workshop on Agent Mediated Electronic Commerce (AmEC-99) (July 1999). Stockholm.

....strategies. In particular, the free riding problem that characterizes equilibrium strategies in other ascending combinatorial auctions [8, 12, 39] is not a problem in BEA. Ascending price auctions can avoid the high cost of information revelation that is required in efficient sealed bid mechanisms [44, 48, 6, 14]. In many interesting problems there is a cost associated with determining the value for a set of items [4] perhaps the bidder in the wireless spectrum auction must determine a new business plan to understand the value of any particular combination of licenses; perhaps the bidder in the ....

....to group minimal CE pces. Proo The special structure provided by NTV prices lows a simplification of the combinatorial constraints in [RD CS ] from which the connection between adjusted CE prices and Vickrey payments is immediate. The full proof is in the Appendix. This explns why Bundle(3) [44, 48] d Ausubel Milgrom s ascending proxy auction can compute the Vickrey outcome when agents are substitutes. In both auctions agents bid prices e approximate NTV prices, d both auctions terminate in competitive equilibrium. The surplus from NTV prices is related to the surplus from Quasi CE ....

David C Parkes. Optimal auction design for agents with hard valuation problems. In Proc. IJCAI-99 Workshop on Agent Mediated Electronic Commerce, pages 206-219, July 1999. Stockholm.

....value for different outcomes. Iterative auctions are useful in these settings, because they allow participants to consider the accuracy to which they should refine their values, and in which parts of the outcome space to focus, all in response to feedback about the bids from other participants [4]. Preference elicitation has previously been considered in the context of iterative combinatorial auctions (e.g. 5, 2] In this paper, we examine the preference elicitation properties of iterative multiattribute auctions, and in particular we consider the effect Harvard College, Cambridge, MA ....

D.C. Parkes. Optimal auction design for agents with hard valuation problems. In Proc. IJCAI-99 Workshop on Agent Mediated Electronic Commerce, pages 206-219, July 1999. Stockhohn.

....ascendingprice combinatorial auction. We have a theoretical proof that the auction terminates with the Vickrey outcome in special cases, and initial experimental results support our conjecture that the auction implements the Vickrey outcome in all cases. The auction extends iBundle (Parkes 1999a; Parkes Ungar 2000b) and builds on the ideas introduced in proxy and adjust (Parkes Ungar 2000c) The goal is to implement the Vickrey outcome with best response agent strategies, i.e. if agents bid in each round for the bundle(s) that maximize their utility. The auction has two distinct phases. The first ....

....terminates with the Vickrey outcome in special cases, and initial experimental results support our conjecture that the auction implements the Vickrey outcome in all cases. The auction extends iBundle (Parkes 1999a; Parkes Ungar 2000b) and builds on the ideas introduced in proxy and adjust (Parkes Ungar 2000c) The goal is to implement the Vickrey outcome with best response agent strategies, i.e. if agents bid in each round for the bundle(s) that maximize their utility. The auction has two distinct phases. The first phase is used to determine the efficient (value maximizing) allocation, while the ....

[Article contains additional citation context not shown here]

Parkes, D. C. 1999b. Optimal auction design for agents with hard valuation problems. In Proc. IJCAI-99 Workshop on Agent Mediated Electronic Commerce. Stockholm.

....agents at the final prices and the auctioneer maximizes its revenue. The iBundle and English auctions terminate in CE. Indeed, a fundamental connection between primal dual optimization theory and competitive equilibrium prices allows optimal auctions to be designed and analyzed (Bertsekas 1990; Parkes Ungar 2000) We introduce Adjust, a procedure to compute minimal CE prices from agents bids after an auction terminates. Minimal CE prices are equilibrium prices that minimize the auctioneer s total revenue from all agents in the optimal allocation. The price paid by each agent with minimal CE prices ....

....compute minimal CE prices from agents bids and prices after an auction terminates. Complementary slackness conditions for appropriate primal and dual formulations of the global resource allocation problem are equivalent to equilibrium conditions between an allocation and prices (Bertsekas 1990; Parkes Ungar 2000) Consider an auction A that terminates in equilibrium, let p i (S) denote the price for bundle S to agent i, and let S i denote the bundle allocated to agent i. In defining a competitive equilibrium we allow price discrimination, with different prices for agents, e.g. p i (S) 6= p j (S) for ....

[Article contains additional citation context not shown here]

Parkes, D. C. 1999a. Optimal auction design for agents with hard valuation problems. In Proc. IJCAI-99 Workshop on Agent Mediated Electronic Commerce. Stockholm.

....because of the computational complexity of strategic behavior. For example, in the FCC broadband spectrum auction, conducted as a set of simultaneous ascending price auctions on spectrum licenses, bids were rarely above minimum ask prices and jump bids were the exception (Cramton 1997) In Parkes Ungar (2000) we present a simple extension to iBundle that makes it robust to strategic manipulation in several interesting problems; we adjust the final prices in iBundle towards Vickrey prices. The Ascending Price Bundle Auction iBundle is an ascending price auction that allows agents to bid on arbitrary ....

Parkes, D. C. 1999a. Optimal auction design for agents with hard valuation problems. In Proc. IJCAI-99 Workshop on Agent Mediated Electronic Commerce. Stockholm.

....has practical significance because it addresses the computational and informational complexity of bundle auctions. For agents, the auction can terminate before agents have revealed (or even computed) their values for every bundle. This can be important when agents have hard valuation problems [15, 16]. Furthermore, the auctioneer only needs to generate explicit prices on a subset of bundles and often solves smaller winner determination problems than in the classic sealed bid bundle auction (the Generalized Vickrey Auction [23] Also relevant, because the winner determination problem remains ....

....ability to tradeoff allocative efficiency for computation and communication cost will be important in practical applications. Iterative auctions can allow more efficient computation than sealed bid auctions when agents have hard valuation problems, and this can lead to higher allocative efficiency [15]. iBundle can generate solutions without information from agents on their value for every bundle. Agents only need to determine the bundles that maximize utility given prices to be able follow a best response bidding strategy, and this can be done with approximate values for bundles. Although ....

Parkes, D. C. Optimal auction design for agents with hard valuation problems. In Proc. 2nd Workshop on Agent Mediated Electronic Commerce (AmEC-99) (July 1999). Stockholm.

No context found.

Parkes, D. 1999b. Optimal auction design for agents with hard valuation problems. In Agent-Mediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence.

No context found.

David C. Parkes. Optimal auction design for agents with hard valuation problems. In Agent-Mediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence, Stockholm, Sweden, 1999b.

No context found.

David C Parkes. Optimal auction design for agents with hard valuation problems. In Agent-Mediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence, Stockholm, Sweden, 1999.

No context found.

D. Parkes, Optimal auction design for agents with hard valuation problems - long version, Proc. IJCAI'99 Workshop on Agent Mediated Electronic Commerce (1999). 69 BIBLIOGRAPHY 70

No context found.

D. C. Parkes. Optimal auction design for agents with hard valuation problems. Agent-Mediated Electronic Commerce Workshop at IJCAI, 1999.

No context found.

D. C. Parkes. Optimal auction design for agents with hard valuation problems. In Agent-Mediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence, Stockholm, Sweden, 1999.

No context found.

David C. Parkes. Optimal auction design for agents with hard valuation problems. In Proceedings of the Agent Mediated Electronic Commerce (IJCAI Workshop), Stockholm, Sweden, pages 206--219, July 1999.

No context found.

Parkes, David C. (2000). "Optimal Auction Design for Agents with Hard Valuation Problems," Agent Mediated Electronic Commerce (IJCAI Workshop)

No context found.

David C Parkes. Optimal auction design for agents with hard valuation problems. In Agent-Mediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence, Stockholm, Sweden, 1999.

No context found.

D. C Parkes. Optimal auction design for agents with hard valuation problems. Agent Mediated Electronic Commerce II: Towards Next-Generation Agent Based Electronic Commerce Systems, LNAI, 1788, 2000.

No context found.

David C Parkes. Optimal auction design for agents with hard valuation problems. IJCAI-99 Agent-Mediated Electronic Commerce Workshop.

No context found.

Parkes, D.C.: Optimal auction design for agents with hard valuation problems. In: Agent-Mediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence, Stockholm, Sweden (1999)

No context found.

David C Parkes. Optimal auction design for agents with hard valuation problems. AMEC Workshop, IJCAI, 1999.

No context found.

David C Parkes. Optimal auction design for agents with hard valuation problems. In Agent-Mediated Electronic Commerce Workshop at the International Joint Conference on Artificial Intelligence, Stockholm, Sweden, 1999.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC