本課程將全面的介紹近年發(fā)展起來的基于神經(jīng)網(wǎng)絡的深度學習技術的基本概念,主要結構,核心方法和關鍵應用。主要內容包括:(1)機器學習和神經(jīng)網(wǎng)絡的基本概念和算法及其背后概率論、線性代數(shù)、優(yōu)化理論相關基礎;(2)深度學習的主流結構、激活函數(shù)、正則化技術,實用算法細節(jié)和應用案例;(3)計算機視覺與白然語言處理技術原理與應用;(4)包括模型壓縮、生成對抗網(wǎng)絡技術在內的新興技術簡介;(5)前沿論文與技術探討。
通過課程的學習,使同學們鞏固基礎數(shù)學及機器學習的基本概念和算法;掌握神經(jīng)網(wǎng)絡基本概念;掌握深度學習中的主要網(wǎng)絡結構的基本概念和相關算法;了解具體應用領域的背景知識、應用相關的深度學習技術;并了解生成對抗網(wǎng)絡、深度神經(jīng)網(wǎng)絡模型壓縮等新興技術。
深度學習是目前人工智能、機器學習領域異常火熱的研究方向,受到了學術界和
工業(yè)界的高度關注,被《麻省理工學院技術評論》(MIT Technology Review)評為2013年十大突破性技術之首。深度學習已經(jīng)在語音識別、圖像識別、自然語言處理等諸多
領域取得了突破性進展,對學術界和工業(yè)界產(chǎn)生了深遠的影響。本課程采用google開源軟件TensorFlow作為深度學習技術實現(xiàn)平臺,講解了全連接神經(jīng)網(wǎng)絡、自編碼器
和多層感知機、卷積神經(jīng)網(wǎng)絡、循環(huán)神經(jīng)網(wǎng)絡等的設計與實現(xiàn),以及網(wǎng)絡訓練過程中的數(shù)據(jù)處理、網(wǎng)絡優(yōu)調與超參數(shù)設計,并介紹深度強化學習和網(wǎng)絡模型的可視化、多
GPU并行與分布式處理技術。通過本課程的學習使學生掌握深度學習技術并應用該技
術解決實際問題,了解應用領域的背景知識。
Deep learning is currently an extremely hot research direction in the field of artificial intelligence and machine learning.It has received great attention from academia and industry.It was rated as one of the top ten breakthrough technologies in 2013 by the MIT Technology Review.The first.Deep learning has made breakthroughs in many fields such as speech recognition,image recognition,and natural language processing,and has had a profound impact on academia and industry.This course uses Google's open source software TensorFlow as the deep learning technology implementation platform,and explains the design and implementation of fully connected neural networks,autoencoders and multilayer perceptrons,convolutional neural networks,recurrent neural networks,etc.,as well as the network training process Data processing,network optimization and hyperparameter design,and introduction of deep reinforcement learning and network model visualization,