人工智能中的深度学习——从机器感知到机器认知-邓力上课讲义.pptxVIP

;;;;;;;;;;;;;;;;Real-time reduction of 16% WER reduction of 10%;;English Conversational Telephone Speech Recognition* Key ingredients:;;;;;;11x11conv,96,/4,pool/2;AlexNet, 8 layers; 1x1 conv, 64 3x3 conv, 64 1x1 conv, 256 1x1 conv, 64 3x3 conv, 64 1x1 conv, 256 1x1 conv, 64 3x3 conv, 64 1x1 conv, 256 1x2 conv, 128, /2 3x3 conv, 128 1x1 conv, 512 1x1 conv, 128 3x3 conv, 128 1x1 conv, 512 1x1 conv, 128 3x3 conv, 128 1x1 conv, 512 1x1 conv, 128 3x3 conv, 128 1x1 conv, 512 1x1 conv, 128 3x3 conv, 128 1x1 conv, 512 1x1 conv, 128 3x3 conv, 128 1x1 conv, 512 1x1 conv, 128 3x3 conv, 128 1x1 conv, 512 1x1 conv, 128 3x3 conv, 128 1x1 conv, 512 1x1 conv, 256, /2;; dim = 100M s: “racing car”;Many applications of Deep Semantic Modeling: Learning semantic relationship between “Source” and “Target”;Language Model;;;;;;;;;;;;;;;;;;;;;;Models for Global Local Attention;;58;;;;;;;;Aux;;Deep Discriminative NN;;;;;;;;;