arXiv subjects

Vineet Abhishek

Publications and source records attributed to Vineet Abhishek.

7 recordsLinked to original sources

arXiv2017

A nonparametric sequential test for online randomized experiments

We propose a nonparametric sequential test that aims to address two practical problems pertinent to online randomized experiments: (i) how to do a hypothesis test for complex metrics; (ii) how to prevent type $1$ error inflation under continuous monitoring. The proposed test does not require knowledge of the underlying probability distribution generating the data. We use the bootstrap to estimate the likelihood for blocks of data followed by mixture sequential probability ratio test. We validate this procedure on data from a major online e-commerce website. We show that the proposed test controls type $1$ error at any time, has good power, is robust to misspecification in the distribution generating the data, and allows quick inference in online randomized experiments.

stat.ML↗

arXiv2017

Fixed and Market Pricing for Cloud Services

We study a model of congestible resources, where pricing and scheduling are intertwined. Motivated by the problem of pricing cloud instances, we model a cloud computing service as linked $GI/GI/\cdot$ queuing systems where the provider chooses to offer a fixed pricing service, a dynamic market based service, or a hybrid of both, where jobs can be preempted in the market-based service. Users (jobs), who are heterogeneous in both the value they place on service and their cost for waiting, then choose between the services offered. Combining insights from auction theory with queuing theory we are able to characterize user equilibrium behavior, and show its insensitivity to the precise market design mechanism used. We then provide theoretical and simulation based evidence suggesting that a fixed price typically, though not always, generates a higher expected revenue than the hybrid system for the provider.

cs.GT↗

arXiv2015

On Bidding with Securities: Risk Aversion and Positive Dependence

DeMarzo et al. (2005) consider auctions in which bids are selected from a completely ordered family of securities whose values are tied to the resource being auctioned. The paper defines a notion of relative steepness of families of securities and shows that a steeper family provides greater expected revenue to the seller. Two assumptions are: the buyers are risk-neutral; the random variables through which values and signals of the buyers are realized are affiliated. We show that this revenue ranking holds for the second price auction in the case of risk-aversion. However, it does not hold if affiliation is relaxed to a less restrictive form of positive dependence, namely first order stochastic dominance (FOSD). We define the relative strong steepness of families of securities and show that it provides a necessary and sufficient condition for comparing two families in the FOSD case. All results extend to the English auction.

cs.GT↗

arXiv2012

On the Incentive to Deviate in Core Selecting Combinatorial Auctions

Recent spectrum auctions in the United Kingdom, and some proposals for future auctions of spectrum in the United States, are based on preliminary price discovery rounds, followed by calculation of final prices for the winning buyers. For example, the prices could be the projection of Vikrey prices onto the core of reported prices. The use of Vikrey prices should lead to more straightforward bidding, but the projection reverses some of the incentive for bidders to report truthfully. Still, we conjecture that the price paid by a winning buyer increases no faster than the bid, as in a first price auction. It would be rather disturbing if the conjecture is false. The conjecture is established for a buyer interacting with disjoint groups of other buyers in a star network setting. It is also shown that for any core-selecting payment rule and any integer w greater than or equal to two, there is a market setting with w winning buyers such that the price paid by some winning buyer increases at least (1-1/w) times as fast as the price bid.

cs.GT↗

arXiv2012

Auctions with a Profit Sharing Contract

We study the problem of selling a resource through an auction mechanism. The winning buyer in turn develops this resource to generate profit. Two forms of payment are considered: charging the winning buyer a one-time payment, or an initial payment plus a profit sharing contract (PSC). We consider a symmetric interdependent values model with risk averse or risk neutral buyers and a risk neutral seller. For the second price auction and the English auction, we show that the seller's expected total revenue from the auction where he also takes a fraction of the positive profit is higher than the expected revenue from the auction with only a one-time payment. Moreover, the seller can generate an even higher expected total revenue if, in addition to taking a fraction of the positive profit, he also takes the same fraction of any loss incurred from developing the resource. Moving beyond simple PSCs, we show that the auction with a PSC from a very general class generates higher expected total revenue than the auction with only a one-time payment. Finally, we show that suitable PSCs provide higher expected total revenue than a one-time payment even when the incentives of the winning buyer to develop the resource must be addressed by the seller.

cs.GT↗

arXiv2010

Revenue Optimal Auction for Single-Minded Buyers

We study the problem of characterizing revenue optimal auctions for single-minded buyers. Each buyer is interested only in a specific bundle of items and has a value for the same. Both his bundle and its value are his private information. The bundles that buyers are interested in and their corresponding values are assumed to be realized from known probability distributions independent across the buyers. We identify revenue optimal auctions with a simple structure, if the conditional distribution of any buyer's valuation is nondecreasing, in the hazard rates ordering of probability distributions, as a function of the bundle the buyer is interested in. The revenue optimal auction is given by the solution of a maximum weight independent set problem. We provide a novel graphical construction of the weights and highlight important properties of the resulting auction.

cs.GT↗

arXiv2010

Efficiency Loss in Revenue Optimal Auctions

We study efficiency loss in Bayesian revenue optimal auctions. We quantify this as the worst case ratio of loss in the realized social welfare to the social welfare that can be realized by an efficient auction. Our focus is on auctions with single-parameter buyers and where buyers' valuation sets are finite. For binary valued single-parameter buyers with independent (not necessarily identically distributed) private valuations, we show that the worst case efficiency loss ratio (ELR) is no worse than it is with only one buyer; moreover, it is at most 1/2. Moving beyond the case of binary valuations but restricting to single item auctions, where buyers' private valuations are independent and identically distributed, we obtain bounds on the worst case ELR as a function of number of buyers, cardinality of buyers' valuation set, and ratio of maximum to minimum possible values that buyers can have for the item.

cs.GT↗