In the stochastic submodular cover problem, the goal is to select a subset of stochastic items of minimum expected cost to cover a submodular function. Solutions in this setting correspond to sequential decision processes that select items one by one "adaptively" (depending on prior observations). While such adaptive solutions achieve the best objective, the inherently sequential nature makes them undesirable in many applications. We ask: how well can solutions with only a few adaptive rounds approximate fully-adaptive solutions? We give nearly tight answers for both independent and correlated settings, proving smooth tradeoffs between the number of adaptive rounds and the solution quality, relative to fully adaptive solutions. Experiments on synthetic and real datasets show qualitative improvements in the solutions as we allow more rounds of adaptivity; in practice, solutions with a few rounds of adaptivity are nearly as good as fully adaptive solutions.
翻译:在随机子模块覆盖问题中,目标是选择一组具有最低预期成本的随机项目来覆盖子模块功能。 此处的解决方案与顺序决定过程相对应, 即“ 适应性” 选择一个项目( 取决于先前的观察结果 ) 。 虽然这些适应性解决方案达到了最佳目标, 但内在的顺序性质使得这些解决方案在许多应用中不可取。 我们问: 仅有几个适应性回合的解决方案能够有多好, 大约是完全适应性的解决方案。 我们为独立和关联的设置给出了近乎紧凑的答案, 证明适应性回合的数量和解决方案质量之间的平衡, 与完全适应性解决方案的相对。 合成和真实数据集实验显示解决方案的质量改进,因为我们允许更多轮的适应性; 实际上, 几轮适应性解决方案几乎和完全适应性解决方案一样好。