首页>
外国专利>
FIXED-POINT TRAINING METHOD FOR DEEP NEURAL NETWORKS BASED ON DYNAMIC FIXED-POINT CONVERSION SCHEME
FIXED-POINT TRAINING METHOD FOR DEEP NEURAL NETWORKS BASED ON DYNAMIC FIXED-POINT CONVERSION SCHEME
展开▼
机译:基于动态不动点转换方案的深层神经网络不动点训练方法
展开▼
页面导航
摘要
著录项
相似文献
摘要
The present disclosure proposes a fixed-point training method and apparatus based on dynamic fixed-point conversion scheme. More specifically, the present disclosure proposes a fixed-point training method for LSTM neural network. According to this method, during the fine-tuning process of the neural network, it uses fixed-point numbers to conduct forward calculation. Accordingly, within several training cycles, the network accuracy may returned to the desired accuracy level under floating point calculation.
展开▼