Test Score Algorithms for Budgeted Stochastic Utility Maximization

Motivated by recent developments in designing algorithms based on individual item scores for solving utility maximization problems, we study the framework of using test scores, defined as a statistic observed performance data, budgeted stochastic problem. We extend an existing scoring mechanism, namely, replication to incorporate heterogeneous costs well values. show that natural greedy algorithm selects items solely their outputs solutions within constant factor optimum class functions satisfying extended diminishing returns property. Our and approximation guarantees assume are noisy estimates certain expected values with respect marginal distributions values, thus making our practical extending previous work assumes noiseless estimates. Moreover, how can be adapted setting which arrive streaming fashion while maintaining same guarantee. present numerical results, synthetic data sets from Academia.StackExchange Q&A forum, score achieve competitiveness some cases better than benchmark requires access value oracle evaluate function Funding: This research was supported, part, Institute Basic Science [Grants IBS-R029-C1, IBS-R029-Y2].