Applying interchangeability techniques to the distributed breakout algorithm
Applying interchangeability techniques to the distributed breakout algorithm
复制标题
DOI：
--
复制
发表时间：
2009-07
期刊：
影响因子：
--
通讯作者：
Xuan Sun;Takuya Matsuzaki;Daisuke Okanohara;Junichi Tsujii
中科院分区：
文献类型：
--
作者：
Xuan Sun;Takuya Matsuzaki;Daisuke Okanohara;Junichi Tsujii
提出了一种感知器式的结构化隐变量模型快速判别训练算法，并分析了其收敛性质。我们的方法扩展了感知器算法的学习任务与潜在的依赖关系，这可能不会被传统的模型。它依赖于对潜在变量的Viterbi解码，结合简单的加法更新。与现有的潜在变量的概率模型相比，我们的方法显著降低了训练成本，但具有相当甚至上级的分类精度。
We propose a perceptron-style algorithm for fast discriminative training of structured latent variable model, and analyzed its convergence properties. Our method extends the perceptron algorithm for the learning task with latent dependencies, which may not be captured by traditional models. It relies on Viterbi decoding over latent variables, combined with simple additive updates. Compared to existing probabilistic models of latent variables, our method lowers the training cost significantly yet with comparable or even superior classification accuracy.