论文标题
等级空间中的动态编程：使用低级别HMM和PCFG的缩放结构化推理
Dynamic Programming in Rank Space: Scaling Structured Inference with Low-Rank HMMs and PCFGs
论文作者
论文摘要
隐藏的马尔可夫模型（HMM）和概率无上下文语法（PCFGS）是广泛使用的结构化模型，两者都可以表示为因素图语法（FGGS），这是一种强大的形式主义，能够描述广泛的模型。最近的研究发现，用于HMM和PCFG的大型状态空间是有益的。但是，对大型状态空间的推论在计算上是要求的，尤其是对于PCFG。为了应对这一挑战，我们利用张量秩分解（aka。\ cpd）来降低FGGS包含的HMMS和PCFG的子集的推理计算复杂性。我们将CPD应用于FGG的因素，然后构建等级空间中定义的新FGG。使用新FGG的推断会产生相同的结果，但是当等级大小小于状态大小时，时间复杂性较低。我们对HMM语言建模和无监督的PCFG解析进行实验，比以前的工作表现更好。我们的代码可在\ url {https://github.com/vpeterv/rankspace-models}上公开获得。
Hidden Markov Models (HMMs) and Probabilistic Context-Free Grammars (PCFGs) are widely used structured models, both of which can be represented as factor graph grammars (FGGs), a powerful formalism capable of describing a wide range of models. Recent research found it beneficial to use large state spaces for HMMs and PCFGs. However, inference with large state spaces is computationally demanding, especially for PCFGs. To tackle this challenge, we leverage tensor rank decomposition (aka.\ CPD) to decrease inference computational complexities for a subset of FGGs subsuming HMMs and PCFGs. We apply CPD on the factors of an FGG and then construct a new FGG defined in the rank space. Inference with the new FGG produces the same result but has a lower time complexity when the rank size is smaller than the state size. We conduct experiments on HMM language modeling and unsupervised PCFG parsing, showing better performance than previous work. Our code is publicly available at \url{https://github.com/VPeterV/RankSpace-Models}.
