AAAI Publications, Thirty-First AAAI Conference on Artificial Intelligence

Font Size: 
Market Pricing for Data Streams
Melika Abolhassani, Hossein Esfandiari, MohammadTaghi Hajiaghayi, Brendan Lucier, Hadi Yami

Last modified: 2017-02-10

Abstract


Internet-enabled marketplaces such as Amazon deal with huge datasets registering transaction of merchandises between lots of buyers and sellers. It is important that algorithms become more time and space efficient as the size of datasets increase. An algorithm that runs in polynomial time may not have a reasonable running time for such large datasets. Here, we study the development of pricing algorithms that are appropriate for use with massive datasets. We especially focus on the streaming setting, the common model for big data analysis. We present an envy-free mechanism for social welfare maximization problem in the streaming setting using O(k2 l) space, where k is the number of different goods and l is the number of available items of each good. We also provide an α-approximation mechanism for revenue maximization in this setting given an α-approximation mechanism for the corresponding offline problem exists. Moreover, we provide mechanisms to approximate the optimum social welfare (or revenue) within 1 – ε factor, in space independent of l which would be favorable in case l is large compared to k. Finally, we present hardness results showing approximation of optimal prices that maximize social welfare (or revenue) in the streaming setting needs Ω(l) space. We achieve our results by developing a powerful sampling technique for bipartite networks. The simplicity of our sampling technique empowers us to maintain the sample over the input sequence. Indeed, one can construct this sample in the distributed setting (a.k.a, MapReduce) and get the same results in two rounds of computations, or one may simply apply this sampling technique to provide faster offline algorithms.

Keywords


Auction; Big Data; Streaming Algorithms, Envy-Free Mechanisms

Full Text: PDF