T1：TensorFlow and Sequence to Sequence Learning

 
讲者: 吴永辉（Google） 
 
Abstract：The tutorial will consists of two parts, the first part on TensorFlow and the second part on sequence to sequence learning. TensorFlow is an open source software library for numerical computation using data flow graphs. It is being widely used both within google and outside of google, for both research and production uses. Since the release of TensorFlow in Nov 2015, it is quickly shaping up as the most popular deep learning platform among the machine learning community. In this tutorial, I will explain the key concepts in TensorFlow. I will conclude the section with a few concrete example machine learning models built using TensorFlow. Sequence to sequence learning represents an import class of machine learning problems. Examples of sequence to sequence learning problems include: automatic speech recognition (ASR), where an audio sequence is being transcribed into a sequence of characters. Machine translation is another example, where a sentence in some source language is being translated into a sentence in another language. Speech synthesis (TTS) is yet another example. In this section, I will give an overview of problems in the three domains (ASR, MT and TTS), popular models for each of the problems, and how a common model architecture, sequence to sequence model with attention, is able to achieve state of the art result in all those problems.（讲义内容为英文，报告语言为中文。）


个人简介：吴永辉，Google Brain团队首席软件工程师。获南京大学理学学士学位，加州大学河滨分校计算机科学博士学位。目前他是Google应用研究团队的技术负责人，主要研究方向是ASR、NMT、TTS、语言建模、排序等。他是Google神经机器翻译项目及Rank Brain项目的主要贡献者之一。在加入Google Brain团队之前，曾与Google的搜索排名小组合作，致力于提高Google网页搜索排名技术。他还是近十年来Google核心排名算法重大里程碑HummingBird的主要贡献者。

Brief introduction：Yonghui Wu is a Principal Software Engineer within the Google Brain team. He is currently a tech leader / manager of an applied research team, whose research focus is on ASR, NMT, TTS, language modeling, ranking and various topics. He is one of the main contributors to the GNMT project, and the RankBrain project, among others. Prior to joining the Google Brain team, he was with the Google's search ranking team, where his focus was to improve Google web search ranking quality. He was one of the main contributors to the HummingBird change, which is the largest rewriting of Google's core ranking algorithm in more than 10 years. Yonghui holds a Ph.D degree in computer science from University of California, Riverside, and a BS degree in CS from Nanjing University.