Bluche T, Kermorvant C, Louradour J. Where to apply dropout in recurrent neural networks for handwriting recognition?[C]// 2015 13th International Conference on Document Analysis and Recognition (ICDAR). IEEE Computer Society, 2015:681-685.
paper to read:Improving neural networks by preventing co-adaption of feature detectors
recurrent neural network regularization
investigation of recurrent neural networks architectures and learning methods for spoken language understanding
Long short-term memory
dropout用在lstm前效果优于中间,优于后面
各层全加dropout不如在最优的位置添加dropout
加入dropout后,权重更加sharper,
LSTM层之后的dropout可以促使不同unit发掘更多的信息,减少相邻unit之间的依赖
LSTM层之前的dropout可以减少相邻像素点之间的依赖
paper to read:Improving neural networks by preventing co-adaption of feature detectors
recurrent neural network regularization
investigation of recurrent neural networks architectures and learning methods for spoken language understanding
Long short-term memory
dropout用在lstm前效果优于中间,优于后面
各层全加dropout不如在最优的位置添加dropout
加入dropout后,权重更加sharper,
LSTM层之后的dropout可以促使不同unit发掘更多的信息,减少相邻unit之间的依赖
LSTM层之前的dropout可以减少相邻像素点之间的依赖