完全是因为adaptive Boosting吸引了我,慢慢积累慢慢成长(有点事,没有更完)
AdaBoosting: 通过顺序的学习一些弱分类器,然后通过加权投票(weight voting)得到最后的预测,每一次迭代后,样本的权重都会更新
先给出paper中出现的符号意思:
Ensemble Neural Relation Extraction with Adaptive Boosting
接下来一一给出公式的解释:
Ensemble Neural Relation Extraction with Adaptive Boosting
公式(4) 是分类器对K个语料进行分类的错误率,公式(3)中的 j 是第j个batch
Ensemble Neural Relation Extraction with Adaptive Boosting
公式(2)
Ensemble Neural Relation Extraction with Adaptive Boosting
公式(6)
Ensemble Neural Relation Extraction with Adaptive Boosting
公式(7)
Ensemble Neural Relation Extraction with Adaptive Boosting
公式(8)

算法伪代码:
Ensemble Neural Relation Extraction with Adaptive Boosting
——————————————————————————————————————————关于ensemble learning:
Ensemble Neural Relation Extraction with Adaptive Boosting

其中bagging学习的过程如下:
Ensemble Neural Relation Extraction with Adaptive Boosting

boosting学习的过程是:
Ensemble Neural Relation Extraction with Adaptive Boosting

参考文献:
https://blog.csdn.net/sinat_26917383/article/details/54667077
https://blog.csdn.net/qq_36330643/article/details/77621232

相关文章: