Beam search is a go-to strategy for decoding neural sequence models. The algorithm can naturally be viewed as a subset optimization problem, albeit one where the corresponding set function does not reflect interactions between candidates. Empirically, this leads to sets often exhibiting high overlap, e.g., strings may differ by only a single word. Yet in use-cases that call for multiple solutions, a diverse or representative set is often desired. To address this issue, we propose a reformulation of beam search, which we call determinantal beam search. Determinantal beam search has a natural relationship to determinantal point processes (DPPs), models over sets that inherently encode intra-set interactions. By posing iterations in beam search as a series of subdeterminant maximization problems, we can turn the algorithm into a diverse subset selection process. In a case study, we use the string subsequence kernel to explicitly encourage n-gram coverage in text generated from a sequence model. We observe that our algorithm offers competitive performance against other diverse set generation strategies in the context of language generation, while providing a more general approach to optimizing for diversity.
翻译:光束搜索是一种解码神经序列模型的战略。 算法可以自然地被视为子优化问题, 尽管相应的设定函数并不反映候选人之间的相互作用。 随机地, 这导致设置往往显示高度重叠, 例如字符串可能只有一个单词不同。 但是在需要多种解决方案的使用情况下, 通常需要一组多样或具有代表性的组合。 为了解决这个问题, 我们建议重新配置波束搜索, 我们称之为决定性波束搜索 。 Didiminatanal 波束搜索与决定因素点进程( DPPs) 有着自然的关系, 即各组的模型之间有着内在编码内部互动的特性。 通过将横线搜索设置成一系列次定义最大化问题, 我们可以将算法转换成一个多样化的子选择过程。 在案例研究中, 我们使用字符串子后序圈来明确鼓励由序列模型生成的正克范围。 我们观察到, 我们的算法在语言生成背景下, 相对于其他不同组合的生成策略具有竞争性性, 提供了一种优化多样性的通用方法 。