有待验证的小Idea

  1. A Nonparametric Information Theoretic Clustering Algorithm和和supervised的方法结合的可能性.
  2. A Conditional Random Field for Multiple-Instance Learning用到样本间的相关性, 类似zhihua zhou的Multi-instance learning by treating instances as non-I.I.D. samples.

你可能感兴趣的:(idea)