Python深度学习算法实践（***）97875**1**693Sudharsan,R**ichandiran东南大学出版社

《Python深度学习算法实践（***）》深入浅出地剖析了深度学习的原理和相关技术。书中使用Python，从基本的数学知识出发，带领读者从零创建一个经典的深度学习网络，使读者在此过程中逐步理解深度学习。书中不仅介绍了深度学习和神经网络的概念、特征等基础知识，对误差反向传播法、卷积神经网络等也有深入讲解，此外还介绍了深度学习相关的实用技巧，自动驾驶、图像生成、强化学习等方面的应用，以及为什么加深层可以提高识别精度等疑难的问题。

• PrefaceSection*nbsp;1“*nbsp;Getting*nbsp;Started*nbsp;with*nbsp;Deep*nbsp;LearningChapter*nbsp;1：*nbsp;Introduction*nbsp;to*nbsp;Deep*nbsp;LearningWhat*nbsp;is*nbsp;deep*nbsp;learning?Biological*nbsp;and*nbsp;artificial*nbsp;neuronsANN*nbsp;and*nbsp;its*nbs*;**yersInput*nbs*;**yerHidden*nbs*;**yerOutput*nbs*;**yerExploring*nbsp;activation*nbsp;functionsThe*nbsp;sigmoid*nbsp;functionThe*nbsp;tanh*nbsp;functionThe*nbsp;Rectified*nbsp;Linear*nbsp;Unit*nbsp;functionThe*nbsp;leaky*nbsp;ReLU*nbsp;functionThe*nbsp;Exponential*nbsp;linear*nbsp;unit*nbsp;functionThe*nbsp;Swish*nbsp;functionThe*nbsp;softmax*nbsp;functionFor***d*nbsp;propagation*nbsp;in*nbsp;ANNHow*nbsp;does*nbsp;ANN*nbsp;learn?Debugging*nbsp;gradient*nbsp;descent*nbsp;with*nbsp;gradient*nbsp;checkingPutting*nbsp;it*nbsp;all*nbsp;togetherBuilding*nbsp;a*nbsp;neural*nbsp;network*nbsp;from*nbsp;scratchSummaryQuestionsFurther*nbsp;readingChapter*nbsp;2：*nbsp;Getting*nbsp;to*nbsp;Know*nbsp;TensorFIowWhat*nbsp;is*nbsp;TensorFIow?Understanding*nbsp;computational*nbsp;graphs*nbsp;and*nbsp;sessionsSessionsVariab***，*nbsp;constants，*nbsp;and*nbsp;***ceholdersVariab***ConstantsPlaceholders*nbsp;and*nbsp;feed*nbsp;dictionariesIntroducing*nbsp;TensorBoardCreating*nbsp;a*nbsp;name*nbsp;scopeHandwritten*nbsp;digit*nbsp;classification*nbsp;using*nbsp;TensorFIowImporting*nbsp;the*nbsp;required*nbsp;librariesLoading*nbsp;the*nbsp;datasetDefining*nbsp;the*nbsp;number*nbsp;of*nbsp;neurons*nbsp;in*nbsp;each*nbs*;**yerDefining*nbsp;***ceholdersFor***d*nbsp;propagationComputing*nbsp;loss*nbsp;and*nbsp;backpropagationComputing*nbsp;accuracyCreating*nbsp;summaryTraining*nbsp;the*nbsp;modelVisualizing*nbsp;graphs*nbsp;in*nbsp;TensorBoardIntroducing*nbsp;eager*nbsp;executionMath*nbsp;operations*nbsp;in*nbsp;TensorFIowTensorFIow*nbsp;2.0*nbsp;and*nbsp;KerasBonjour*nbsp;KerasDefining*nbsp;the*nbsp;modelDefining*nbsp;a*nbsp;sequential*nbsp;modelDefining*nbsp;a*nbsp;functional*nbsp;modelCompiling*nbsp;the*nbsp;modelTraining*nbsp;the*nbsp;modelEvaluating*nbsp;the*nbsp;modelMNIST*nbsp;digit*nbsp;classification*nbsp;using*nbsp;TensorFIow*nbsp;2.0Should*nbsp;we*nbsp;use*nbsp;Keras*nbsp;or*nbsp;TensorFIow?SummaryQuestionsFurther*nbsp;readingSection*nbsp;2：*nbsp;Fundamental*nbsp;Deep*nbsp;Learning*nbsp;AlgorithmsChapter*nbsp;3：*nbsp;Gradient*nbsp;Descent*nbsp;and*nbsp;Its*nbsp;VariantsDemystifying*nbsp;gradient*nbsp;descentPerforming*nbsp;gradient*nbsp;descent*nbsp;in*nbsp;regression*nbsp;“Importing*nbsp;the*nbsp;librariesPreparing*nbsp;the*nbsp;datasetDefining*nbsp;the*nbsp;loss*nbsp;functionComputing*nbsp;the*nbsp;gradients*nbsp;of*nbsp;the*nbsp;loss*nbsp;functionUpdating*nbsp;the*nbsp;model*nbsp;parametersGradient*nbsp;descent*nbsp;versus*nbsp;stochastic*nbsp;gradient*nbsp;descentMomentum-based*nbsp;gradient*nbsp;descentGradient*nbsp;descent*nbsp;with*nbsp;momentumNesterov*nbsp;accelerated*nbsp;gradientAdaptive*nbsp;methods*nbsp;of*nbsp;gradient*nbsp;descentSetting*nbsp;a*nbsp;learning*nbsp;rate*nbsp;adaptively*nbsp;using*nbsp;AdagradDoing*nbsp;away*nbsp;with*nbsp;the*nbsp;learning*nbsp;rate*nbsp;using*nbsp;AdadeltaOvercoming*nbsp;the*nbsp;limitations*nbsp;of*nbsp;Adagrad*nbsp;using*nbsp;RMSPropAdaptive*nbsp;moment*nbsp;estimationAdamax*nbsp;-*nbsp;Adam*nbsp;based*nbsp;on*nbsp;infinity-normAdaptive*nbsp;moment*nbsp;estimation*nbsp;with*nbsp;AMSGrad**Section*nbsp;3*nbsp;Advanced*nbsp;Deep*nbsp;Learning*nbsp;Algorithms