Abstract

Bidders in online multi-unit combinatorial auctions face the acute problem of estimating the valuations of an immense number of packages. Can the seller guide the bidders to avoid placing bids that are too high or too low? In the single unit case, fast methods are now available for incrementally computing, for each package at each time instant, the recommended lower bound (Deadness Level) and upper bound (Winning Level) on the next bid. But when there are multiple units of items, it becomes difficult to compute the Deadness Level of a package accurately. An upper bound on this quantity can be derived however, and a bid that stays within this bound and the Winning Level is “safe”, in the sense that it is not wasted and has the potential to become a winning bid. What is now needed is an incremental procedure for speeding up the computation of this bound.

Share

COinS
 

Providing Information Feedback to Bidders in Online Multi-unit Combinatorial Auctions

Bidders in online multi-unit combinatorial auctions face the acute problem of estimating the valuations of an immense number of packages. Can the seller guide the bidders to avoid placing bids that are too high or too low? In the single unit case, fast methods are now available for incrementally computing, for each package at each time instant, the recommended lower bound (Deadness Level) and upper bound (Winning Level) on the next bid. But when there are multiple units of items, it becomes difficult to compute the Deadness Level of a package accurately. An upper bound on this quantity can be derived however, and a bid that stays within this bound and the Winning Level is “safe”, in the sense that it is not wasted and has the potential to become a winning bid. What is now needed is an incremental procedure for speeding up the computation of this bound.