Benjamin E. Birnbaum

On-line bipartite matching made simple

We examine the classic on-line bipartite matching problem studied by Karp, Vazirani, and Vazirani [8] and provide a simple proof of their result that the Ranking algorithm for this problem achieves a competitive ratio of 1 -- 1/e.

Improving community health worker performance through automated SMS

Community health workers (CHWs) have been shown to be an effective and powerful intervention for improving community health. Routine visits, for example, can lower maternal and neonatal mortality rates. Despite these benefits, many challenges, including supervision and support, make CHW programs difficult to maintain. An increasing number of mHealth projects are providing CHWs with mobile phones to support their work, which opens up opportunities for real-time supervision of the program. Taking advantage of this potential, we evaluated the impact of SMS reminders to improve the promptness of routine CHW visits, first in a pilot study in Dodoma, Tanzania, followed by two larger studies with 87 CHWs in Dar es Salaam, Tanzania. The first Dar es Salaam study evaluated an escalating reminder system that sent SMS reminders directly to the CHW before notifying the CHWs supervisor after several overdue days. The reminders resulted in an 86% reduction in the average number of days a CHWs clients were overdue (9.7 to 1.4 days), with only a small number of cases ever escalating to the supervisor. However, when the step of escalating to the supervisor was removed in the second study, CHW performance significantly decreased.

Improved Approximation Algorithms for Budgeted Allocations

We provide a 3/2-approximation algorithm for an offline budgetedallocations problem with applications to sponsored search auctions.This an improvement over the e/(e- 1)approximation of Andelman and Mansour [1] and thee/(e- 1) - eapproximation (fore≅ 0.0001) of Feige and Vondrak [2] for themore general Maximum Submodular Welfare (SMW) problem. For aspecial case of our problem, we improve this ratio to

Distributed algorithms via gradient descent for fisher markets

\sqrt{2}

An Improved Analysis for a Greedy Remote-Clique Algorithm Using Factor-Revealing LPs

.We also show that the problem is APX-hard.

Convergence of Local Dynamics to Balanced Outcomes in Exchange Networks

Designing distributed algorithms that converge quickly to an equilibrium is one of the foremost research goals in algorithmic game theory, and convex programs have played a crucial role in the design of algorithms for Fisher markets. In this paper we shed new light on both aspects for Fisher markets with linear and spending constraint utilities. We show fast convergence of the Proportional Response dynamics recently introduced by Wu and Zhang. The convergence is obtained from a new perspective: we show that the Proportional Response dynamics is equivalent to a gradient descent algorithm (with respect to a Bregman divergence instead of euclidean distance) on a convex program that captures the equilibria for linear utilities. We further show that the convex program program easily extends to the case of spending constraint utilities, thus resolving an open question raised by Vazirani. This also gives a way to extend the Proportional Response dynamics to spending constraint utilties. We also prove a technical result that is interesting in its own right: that the gradient descent algorithm based on a Bregman divergence converges with rate O(1/t) under a condition that is weaker than having Lipschitz continuous gradient (which is the usual assumption in the optimization literature for obtaining the same rate).

Automated quality control for mobile data collection

Abstract Given a positive integer k and a complete graph with non-negative edge weights satisfying the triangle inequality, the remote-clique problem is to find a subset of k vertices having a maximum-weight induced subgraph. A greedy algorithm for the problem has been shown to have an approximation ratio of 4, but this analysis was not shown to be tight. In this paper, we use the technique of factor-revealing linear programs to show that the greedy algorithm actually achieves an approximation ratio of 2, which is tight.

Using behavioral data to identify interviewer fabrication in surveys

Bargaining games on exchange networks have been studied by both economists and sociologists. A Balanced Outcome for such a game is an equilibrium concept that combines notions of stability and fairness. In a recent paper, Kleinberg and Tardos introduced balanced outcomes to the computer science community and provided a polynomial-time algorithm to compute the set of such outcomes. Their work left open a pertinent question: are there natural, local dynamics that converge quickly to a balanced outcome? In this paper, we provide a partial answer to this question by showing that simple edge-balancing dynamics converge to a balanced outcome whenever one exists.

On Revenue Maximization in Second-Price Ad Auctions

Systematic interviewer error is a potential issue in any health survey, and it can be especially pernicious in low- and middle-income countries, where survey teams may face problems of limited supervision, chaotic environments, language barriers, and low literacy. Survey teams in such environments could benefit from software that leverages mobile data collection tools to provide solutions for automated data quality control. As a first step in the creation of such software, we investigate and test several algorithms that find anomalous patterns in data. We validate the algorithms using one labeled data set and two unlabeled data sets from two community outreach programs in East Africa. In the labeled set, some of the data is known to be fabricated and some is believed to be relatively accurate. The unlabeled sets are from actual field operations. We demonstrate the feasibility of tools for automated data quality control by showing that the algorithms detect the fake data in the labeled set with a high sensitivity and specificity, and that they detect compelling anomalies in the unlabeled sets.

An improved analysis for a greedy remote-clique algorithm using factor-revealing LPs

Surveys conducted by human interviewers are one of the principal means of gathering data from all over the world, but the quality of this data can be threatened by interviewer fabrication. In this paper, we investigate a new approach to detecting interviewer fabrication automatically. We instrument electronic data collection software to record logs of low-level behavioral data and show that supervised classification, when applied to features extracted from these logs, can identify interviewer fabrication with an accuracy of up to 96%. We show that even when interviewers know that our approach is being used, have some knowledge of how it works, and are incentivized to avoid detection, it can still achieve an accuracy of 86%. We also demonstrate the robustness of our approach to a moderate amount of label noise and provide practical recommendations, based on empirical evidence, on how much data is needed for our approach to be effective.