This preview shows page 1. Sign up to view the full content.
Unformatted text preview: es and every agent will
receive a bundle in the provisional allocation. In fact, iBundle is provably e cient with
myopic bestresponse agent strategies. 3.2.3 Communication Costs: Distributed Methods
Shoham & Tennenholtz ST01 explore the communication complexity of computing simple
functions within an auctionbased algorithm i.e., with selfinterested agents with private
information. Essentially, the authors propose a method to compute solutions to simple
functions with minimal communication complexity. Communication from the auctioneer
to the agents is free in their model, while communication from agents to the auctioneer is
costly. Given this, Shoham & Tennenholtz essentially provide incentive schemes so that
each agent i announces its value vi by sending a single bit to the mechanism whenever the
price in an auction is equal to this value. Max and min functions can be computed with a
single bit from agents, and any function over n agents can be computed in n bits, which
is the lower informationtheoretic bound.
Feigenbaum et al. FPS00 investigate costsharing algorithms for multicast transmission, in which a population of consumers sit on the nodes of a multicast tree. Each user
has a value to receive a shared information stream, such as a lm, and each arc in the
multicast tree has an associated cost. The mechanism design problem is to implement the
multicast solution that maximizes total user value minus total network cost, and shares
the cost across endusers. Noting that budgetbalance, e ciency, and strategyproofness
are impossibility in combination the authors compare the computational properties of a
VickreyClarkeGroves marginal cost MC mechanism e cient and strategyproof and
a Shapley value SH mechanism budgetbalanced and coalitional strategyproof.
A distributed algorithm is developed for MC, in which intermediate nodes in the tree
receive messages, perform some computation, and send messages to their neighbors. The
method, a bottomup followed by a topdown traversal of the tree, computes the solution
to MC with minimal communication complexity, with exactly two messages sent per link.
In comparison, there is no method for the SH mechanism with e cient communication
complexity. All solutions are maximal, and require as many messages per link as in a naive
80 centralized approach. Hence, communication complexity considerations lead to a strong
preference for the MC mechanism, which is not budgetbalanced. The study leaves many
interesting open questions; e.g. are all budgetbalanced solutions maximal, and what are
the gametheoretic properties of alternative strategyproof minimal mechanisms?
The economic literature contains a few notable models of the e ect of limited communication and agent boundedrationality in mechanism design, and in systems of distributed
decision making and information processing. This work is relevant here, given the focus
in my dissertation on computational mechanism design and in particular on the costs of
complete info...
View
Full
Document
This document was uploaded on 03/19/2014 for the course COMP CS286r at Harvard.
 Fall '13
 DavidParkes

Click to edit the document details