random walk theory

Inspired by the theory of simple random walk, we propose a unified framework for semi-supervised learning, i.e. multilevel random walk ( MRW). MRW can produce numerous multi-class label propagation algorithms by constructing various component random walk models based on prior knowledge.

在简单随机游走理论的启发下，我们提出一个统一的半监督学习框架&多层随机游走，通过构建各种组件级的随机游走模型，可以产生一系列对应的多分类标签传播算法。

We extended Torrey's theory, based on the model of random walk to nearest neighbor sites for homonuclear dipolar interaction, to the case that both homo-and heteronuclear interactions are existent.

我们把Torrey同核偶极作用近邻无规扩散行步理论推广到同核、异核偶极作用并存的情况；

Two algorithms, graph cut and random walk based on graph theory, were applied to motion segmentation.

将两种基于图论的算法图切割与随机游走应用于运动对象的分割.