论文部分内容阅读
Recurrent neural network language models (RNNLMs) have been applied in a wide range of research fields, including nature language processing and speech recognition. One challenge in training RNNLMs is the heavy computational cost of the crucial back-propa