1.1标量、向量、张量之间的联系 1 [填空题] _________________________________
1.2张量与矩阵的区别? 1 [填空题] _________________________________
1.3矩阵和向量相乘结果 1 [填空题] _________________________________
1.4向量和矩阵的范数归纳 1 [填空题] _________________________________
1.5如何判断一个矩阵为正定? 2 [填空题] _________________________________
1.6导数偏导计算 3 [填空题]
_________________________________
1.7导数和偏导数有什么区别? 3 [填空题] _________________________________
1.8特征值分解与特征向量 3 [填空题] _________________________________
1.9奇异值与特征值有什么关系? 4 [填空题] _________________________________
1.10机器学习为什么要使用概率? 4 [填空题] _________________________________
1.11变量与随机变量有什么区别? 4 [填空题] _________________________________
1.12常见概率分布? 5 [填空题] _________________________________
1.13举例理解条件概率 9 [填空题]
_________________________________
1.14联合概率与边缘概率联系区别? 10 [填空题] _________________________________
1.15条件概率的链式法则 10 [填空题] _________________________________
1.16独立性和条件独立性 11 [填空题]
_________________________________
1.17期望、方差、协方差、相关系数总结 11 [填空题] * _________________________________
2.1 各种常见算法图示 14 [填空题]
_________________________________
2.2监督学习、非监督学习、半监督学习、弱监督学习? 15 [填空题] _________________________________
2.3 监督学习有哪些步骤 16 [填空题] _________________________________
2.4 多实例学习? 17 [填空题]
_________________________________
2.5 分类网络和回归的区别? 17 [填空题] _________________________________
2.6 什么是神经网络? 17 [填空题] _________________________________
2.7 常用分类算法的优缺点? 18 [填空题]
_________________________________
2.8 正确率能很好的评估分类算法吗? 20 [填空题] _________________________________
2.9 分类算法的评估方法? 20 [填空题] _________________________________
2.10 什么样的分类器是最好的? 22 [填空题] _________________________________
2.11大数据与深度学习的关系 22 [填空题] _________________________________
2.12 理解局部最优与全局最优 23 [填空题] _________________________________
2.13 理解逻辑回归 24 [填空题]
_________________________________
2.14 逻辑回归与朴素贝叶斯有什么区别? 24 [填空题] _________________________________
2.15 为什么需要代价函数? 25 [填空题] _________________________________
2.16 代价函数作用原理 25 [填空题] _________________________________
2.17 为什么代价函数要非负? 26 [填空题] _________________________________
2.18 常见代价函数? 26 [填空题]
_________________________________
2.19为什么用交叉熵代替二次代价函数 28 [填空题]
_________________________________
2.20 什么是损失函数? 28 [填空题] _________________________________
2.21 常见的损失函数 28 [填空题]
_________________________________
2.22 逻辑回归为什么使用对数损失函数? 30 [填空题] _________________________________
0.00 对数损失函数是如何度量损失的? 31 [填空题] _________________________________
2.23 机器学习中为什么需要梯度下降? 32 [填空题] _________________________________
2.24 梯度下降法缺点? 32 [填空题] _________________________________
2.25 梯度下降法直观理解? 32 [填空题] _________________________________
2.23 梯度下降法算法描述? 33 [填空题] _________________________________
2.24 如何对梯度下降法进行调优? 35 [填空题] _________________________________
2.25 随机梯度和批量梯度区别? 35 [填空题] _________________________________
2.26 各种梯度下降法性能比较 37 [填空题] _________________________________
2.27计算图的导数计算图解? 37 [填空题] _________________________________
2.28 线性判别分析（LDA）思想总结 39 [填空题] _________________________________
2.29 图解LDA核心思想 39 [填空题] _________________________________
2.30 二类LDA算法原理? 40 [填空题] _________________________________
2.30 LDA算法流程总结? 41 [填空题] _________________________________
2.31 LDA和PCA区别? 41 [填空题] _________________________________
2.32 LDA优缺点? 41 [填空题]
_________________________________
2.33 主成分分析（PCA）思想总结 42 [填空题] _________________________________
2.34 图解PCA核心思想 42 [填空题] _________________________________
2.35 PCA算法推理 43 [填空题] _________________________________
2.36 PCA算法流程总结 44 [填空题] _________________________________
2.37 PCA算法主要优缺点 45 [填空题] _________________________________
2.38 降维的必要性及目的 45 [填空题] _________________________________
2.39 KPCA与PCA的区别? 46 [填空题] _________________________________
2.40模型评估 47 [填空题]
_________________________________
2.40.1模型评估常用方法? 47 [填空题] _________________________________
2.40.2 经验误差与泛化误差 47 [填空题] _________________________________
2.40.3 图解欠拟合、过拟合 48 [填空题] _________________________________
2.40.4 如何解决过拟合与欠拟合? 49 [填空题] _________________________________
2.40.5 交叉验证的主要作用? 50 [填空题] _________________________________
2.40.6 k折交叉验证? 50 [填空题] _________________________________
2.40.7 混淆矩阵 50 [填空题]
_________________________________
2.40.8 错误率及精度 51 [填空题] _________________________________
2.40.9 查准率与查全率 51 [填空题] _________________________________
2.40.10 ROC与AUC 52 [填空题] _________________________________
2.40.11如何画ROC曲线? 53 [填空题]
_________________________________
2.40.12如何计算TPR，FPR? 54 [填空题] _________________________________
2.40.13如何计算Auc? 56 [填空题]
_________________________________
2.40.14为什么使用Roc和Auc评价分类器? 56 [填空题] _________________________________
2.40.15 直观理解AUC 56 [填空题]
_________________________________
2.40.16 代价敏感错误率与代价曲线 57 [填空题] _________________________________
2.40.17 模型有哪些比较检验方法 59 [填空题] _________________________________
2.40.18 偏差与方差 59 [填空题] _________________________________
2.40.19为什么使用标准差? 60 [填空题] _________________________________
2.40.20 点估计思想 61 [填空题] _________________________________
2.40.21 点估计优良性原则? 61 [填空题]
_________________________________
2.40.22点估计、区间估计、中心极限定理之间的联系? 62 [填空题] _________________________________
2.40.23 类别不平衡产生原因? 62 [填空题] _________________________________
2.40.24 常见的类别不平衡问题解决方法 62 [填空题] _________________________________
2.41 决策树 64 [填空题]
_________________________________
2.41.1 决策树的基本原理 64 [填空题] _________________________________
2.41.2 决策树的三要素? 64 [填空题] _________________________________
2.41.3 决策树学习基本算法 65 [填空题] _________________________________
2.41.4 决策树算法优缺点 65 [填空题] _________________________________
2.40.5熵的概念以及理解 66 [填空题] _________________________________
2.40.6 信息增益的理解 66 [填空题] _________________________________
2.40.7 剪枝处理的作用及策略? 67 [填空题] _________________________________
2.41 支持向量机 67 [填空题]
_________________________________
2.41.1 什么是支持向量机 67 [填空题] _________________________________
2.25.2 支持向量机解决的问题? 68 [填空题] _________________________________
2.25.2 核函数作用? 69 [填空题] _________________________________
2.25.3 对偶问题 69 [填空题]
_________________________________
2.25.4 理解支持向量回归 69 [填空题] _________________________________
2.25.5 理解SVM（核函数） 69 [填空题] _________________________________
2.25.6 常见的核函数有哪些? 69 [填空题] _________________________________
2.25.6 软间隔与正则化 73 [填空题] _________________________________
2.25.7 SVM主要特点及缺点? 73 [填空题] _________________________________
2.26 贝叶斯 74 [填空题]
_________________________________
2.26.1 图解极大似然估计 74 [填空题]
_________________________________
2.26.2 朴素贝叶斯分类器和一般的贝叶斯分类器有什么区别? 76 [填空题] _________________________________
2.26.4 朴素与半朴素贝叶斯分类器 76 [填空题] _________________________________
2.26.5 贝叶斯网三种典型结构 76 [填空题] _________________________________
2.26.6 什么是贝叶斯错误率 76 [填空题]
_________________________________
2.26.7 什么是贝叶斯最优错误率 76 [填空题] _________________________________
2.27 EM算法解决问题及实现流程 76 [填空题] _________________________________
2.28 为什么会产生维数灾难? 78 [填空题] _________________________________
2.29怎样避免维数灾难 82 [填空题]
_________________________________
2.30聚类和降维有什么区别与联系? 82 [填空题] _________________________________
2.31 GBDT和随机森林的区别 83 [填空题] _________________________________
2.32 四种聚类方法之比较 84 [填空题] * _________________________________
3.1基本概念 88 [填空题]
_________________________________
3.1.1神经网络组成? 88 [填空题]
_________________________________
3.1.2神经网络有哪些常用模型结构? 90 [填空题] _________________________________
3.1.3如何选择深度学习开发平台? 92 [填空题] _________________________________
3.1.4为什么使用深层表示 92 [填空题] _________________________________
3.1.5为什么深层神经网络难以训练? 93 [填空题] _________________________________
3.1.6深度学习和机器学习有什么不同 94 [填空题] _________________________________
3.2 网络操作与计算 95 [填空题] _________________________________
3.2.1前向传播与反向传播? 95 [填空题] _________________________________
3.2.2如何计算神经网络的输出? 97 [填空题] _________________________________
3.2.3如何计算卷积神经网络输出值? 98 [填空题] _________________________________
3.2.4如何计算Pooling层输出值输出值? 101 [填空题] _________________________________
3.2.5实例理解反向传播 102 [填空题] _________________________________
3.3超参数 105 [填空题]
_________________________________
3.3.1什么是超参数? 105 [填空题]
_________________________________
3.3.2如何寻找超参数的最优值? 105 [填空题] _________________________________
3.3.3超参数搜索一般过程? 106 [填空题] _________________________________
3.4激活函数 106 [填空题]
_________________________________
3.4.1为什么需要非线性激活函数? 106 [填空题] _________________________________
3.4.2常见的激活函数及图像 107 [填空题] _________________________________
3.4.3 常见激活函数的导数计算? 109 [填空题] _________________________________
3.4.4激活函数有哪些性质? 110 [填空题] _________________________________
3.4.5 如何选择激活函数? 110 [填空题] _________________________________
3.4.6使用ReLu激活函数的优点? 111 [填空题] _________________________________
3.4.7什么时候可以用线性激活函数? 111 [填空题]
_________________________________
3.4.8怎样理解Relu（<0时）是非线性激活函数? 111 [填空题] _________________________________
3.4.9 Softmax函数如何应用于多分类? 112 [填空题] _________________________________
3.5 Batch_Size 113 [填空题]
_________________________________
3.5.1为什么需要Batch_Size? 113 [填空题] _________________________________
3.5.2 Batch_Size值的选择 114 [填空题]
_________________________________
3.5.3在合理范围内，增大 Batch_Size 有何好处? 114 [填空题] _________________________________
3.5.4盲目增大 Batch_Size 有何坏处? 114 [填空题]
_________________________________
3.5.5调节 Batch_Size 对训练效果影响到底如何? 114 [填空题] _________________________________
3.6 归一化 115 [填空题]
_________________________________
3.6.1归一化含义? 115 [填空题] _________________________________
3.6.2为什么要归一化 115 [填空题]
_________________________________
3.6.3为什么归一化能提高求解最优解速度? 115 [填空题] _________________________________
3.6.4 3D图解未归一化 116 [填空题] _________________________________
3.6.5归一化有哪些类型? 117 [填空题] _________________________________
3.6.6局部响应归一化作用 117 [填空题] _________________________________
3.6.7理解局部响应归一化公式 117 [填空题]
_________________________________
3.6.8什么是批归一化（Batch Normalization） 118 [填空题] _________________________________
3.6.9批归一化（BN）算法的优点 119 [填空题] _________________________________
3.6.10批归一化（BN）算法流程 119 [填空题] _________________________________
3.6.11批归一化和群组归一化 120 [填空题]
_________________________________
3.6.12 Weight Normalization和Batch Normalization 120 [填空题] _________________________________
3.7 预训练与微调(fine tuning) 121 [填空题]
_________________________________
3.7.1为什么无监督预训练可以帮助深度学习? 121 [填空题] _________________________________
3.7.2什么是模型微调fine tuning 121 [填空题] _________________________________
3.7.3微调时候网络参数是否更新? 122 [填空题] _________________________________
3.7.4 fine-tuning模型的三种状态 122 [填空题] _________________________________
3.8权重偏差初始化 122 [填空题] _________________________________
3.8.1 全都初始化为0 122 [填空题] _________________________________
3.8.2 全都初始化为同样的值 123 [填空题] _________________________________
3.8.3 初始化为小的随机数 124 [填空题] _________________________________
3.8.4用1/sqrt(n)校准方差 125 [填空题]
_________________________________
3.8.5稀疏初始化(Sparse Initialazation) 125 [填空题] _________________________________
3.8.6初始化偏差 125 [填空题] _________________________________
3.9 Softmax 126 [填空题]
_________________________________
3.9.1 Softmax定义及作用 126 [填空题] _________________________________
3.9.2 Softmax推导 126 [填空题]
_________________________________
3.10 理解One Hot Encodeing原理及作用? 126 [填空题] _________________________________
3.11 常用的优化器有哪些 127 [填空题] _________________________________
3.12 Dropout 系列问题 128 [填空题] _________________________________
3.12.1 dropout率的选择 128 [填空题] _________________________________
3.27 Padding 系列问题 128 [填空题] * _________________________________
4.1LetNet5 129 [填空题]
_________________________________
4.1.1模型结构 129 [填空题]
_________________________________
4.1.2模型结构 129 [填空题]
_________________________________
4.1.3 模型特性 131 [填空题]
_________________________________
4.2 AlexNet 131 [填空题]
_________________________________
4.2.1 模型结构 131 [填空题]
_________________________________
4.2.2模型解读 131 [填空题]
_________________________________
4.2.3模型特性 135 [填空题]
_________________________________
4.3 可视化ZFNet-解卷积 135 [填空题] _________________________________
4.3.1 基本的思想及其过程 135 [填空题] _________________________________
4.3.2 卷积与解卷积 136 [填空题] _________________________________
4.3.3卷积可视化 137 [填空题]
_________________________________
4.3.4 ZFNe和AlexNet比较 139 [填空题] _________________________________
4.4 VGG 140 [填空题]
_________________________________
4.1.1 模型结构 140 [填空题]
_________________________________
4.1.2 模型特点 140 [填空题]
_________________________________
4.5 Network in Network 141 [填空题] _________________________________
4.5.1 模型结构 141 [填空题]
_________________________________
4.5.2 模型创新点 141 [填空题] _________________________________
4.6 GoogleNet 143 [填空题]
_________________________________
4.6.1 模型结构 143 [填空题]
_________________________________
4.6.2 Inception 结构 145 [填空题] _________________________________
4.6.3 模型层次关系 146 [填空题] _________________________________
4.7 Inception 系列 148 [填空题] _________________________________
4.7.1 Inception v1 148 [填空题] _________________________________
4.7.2 Inception v2 150 [填空题] _________________________________
4.7.3 Inception v3 153 [填空题] _________________________________
4.7.4 Inception V4 155 [填空题] _________________________________
4.7.5 Inception-ResNet-v2 157 [填空题] _________________________________
4.8 ResNet及其变体 158 [填空题] _________________________________
4.8.1重新审视ResNet 159 [填空题] _________________________________
4.8.2残差块 160 [填空题]
_________________________________
4.8.3 ResNet架构 162 [填空题] _________________________________
4.8.4残差块的变体 162 [填空题] _________________________________
4.8.5 ResNeXt 162 [填空题]
_________________________________
4.8.6 Densely Connected CNN 164 [填空题] _________________________________
4.8.7 ResNet作为小型网络的组合 165 [填空题] _________________________________
4.8.8 ResNet中路径的特点 166 [填空题]
_________________________________
4.9为什么现在的CNN模型都是在GoogleNet、VGGNet或者AlexNet上调整的? 167 [填空题] *
_________________________________
5.1 卷积神经网络的组成层 170 [填空题] _________________________________
5.2 卷积如何检测边缘信息? 171 [填空题] _________________________________
5.2 卷积的几个基本定义? 174 [填空题] _________________________________
5.2.1卷积核大小 174 [填空题] _________________________________
5.2.2卷积核的步长 174 [填空题] _________________________________
5.2.3边缘填充 174 [填空题]
_________________________________
5.2.4输入和输出通道 174 [填空题] _________________________________
5.3 卷积网络类型分类? 174 [填空题] _________________________________
5.3.1普通卷积 174 [填空题]
_________________________________
5.3.2扩张卷积 175 [填空题]
_________________________________
5.3.3转置卷积 176 [填空题]
_________________________________
5.3.4可分离卷积 177 [填空题]
_________________________________
5.3 图解12种不同类型的2D卷积? 178 [填空题] _________________________________
5.4 2D卷积与3D卷积有什么区别? 181 [填空题] _________________________________
5.4.1 2D 卷积 181 [填空题]
_________________________________
5.4.2 3D卷积 182 [填空题]
_________________________________
5.5 有哪些池化方法? 183 [填空题]
_________________________________
5.5.1一般池化（General Pooling） 183 [填空题] _________________________________
5.5.2重叠池化（OverlappingPooling） 184 [填空题] _________________________________
5.5.3空金字塔池化（Spatial Pyramid Pooling） 184 [填空题] _________________________________
5.6 1x1卷积作用? 186 [填空题]
_________________________________
5.7卷积层和池化层有什么区别? 187 [填空题] _________________________________
5.8卷积核一定越大越好? 189 [填空题] _________________________________
5.9每层卷积只能用一种尺寸的卷积核? 189 [填空题] _________________________________
5.10怎样才能减少卷积层参数量? 190 [填空题]
_________________________________
5.11卷积操作时必须同时考虑通道和区域吗? 191 [填空题] _________________________________
5.12采用宽卷积的好处有什么? 192 [填空题] _________________________________
5.12.1窄卷积和宽卷积 192 [填空题] _________________________________
5.12.2 为什么采用宽卷积? 192 [填空题]
_________________________________
5.13卷积层输出的深度与哪个部件的个数相同? 192 [填空题] _________________________________
5.14 如何得到卷积层输出的深度? 193 [填空题]
_________________________________
5.15激活函数通常放在卷积神经网络的那个操作之后? 194 [填空题] _________________________________
5.16 如何理解最大池化层有几分缩小? 194 [填空题] _________________________________
5.17理解图像卷积与反卷积 194 [填空题] _________________________________
5.17.1图像卷积 194 [填空题]
_________________________________
5.17.2图像反卷积 196 [填空题]
_________________________________
5.18不同卷积后图像大小计算? 198 [填空题] _________________________________
5.18.1 类型划分 198 [填空题]
_________________________________
5.18.2 计算公式 199 [填空题]
_________________________________
5.19 步长、填充大小与输入输出关系总结? 199 [填空题] _________________________________
5.19.1没有0填充，单位步长 200 [填空题] _________________________________
5.19.2零填充，单位步长 200 [填空题] _________________________________
5.19.3不填充，非单位步长 202 [填空题] _________________________________
5.19.4零填充，非单位步长 202 [填空题] _________________________________
5.20 理解反卷积和棋盘效应 204 [填空题] _________________________________
5.20.1为什么出现棋盘现象? 204 [填空题]
_________________________________
5.20.2 有哪些方法可以避免棋盘效应? 205 [填空题] _________________________________
5.21 CNN主要的计算瓶颈? 207 [填空题]
_________________________________
5.22 CNN的参数经验设置 207 [填空题] _________________________________
5.23 提高泛化能力的方法总结 208 [填空题] _________________________________
5.23.1 主要方法 208 [填空题]
_________________________________
5.23.2 实验证明 208 [填空题]
_________________________________
5.24 CNN在CV与NLP领域运用的联系与区别? 213 [填空题] _________________________________
5.24.1联系 213 [填空题]
_________________________________
5.24.2区别 213 [填空题]
_________________________________
5.25 CNN凸显共性的手段? 213 [填空题] _________________________________
5.25.1 局部连接 213 [填空题]
_________________________________
5.25.2 权值共享 214 [填空题]
_________________________________
5.25.3 池化操作 215 [填空题]
_________________________________
5.26 全卷积与Local-Conv的异同点 215 [填空题] _________________________________
5.27 举例理解Local-Conv的作用 215 [填空题] _________________________________
5.28 简述卷积神经网络进化史 216 [填空题] * _________________________________
6.1 RNNs和FNNs有什么区别? 218 [填空题] _________________________________
6.2 RNNs典型特点? 218 [填空题] _________________________________
6.3 RNNs能干什么? 219 [填空题]
_________________________________
6.4 RNNs在NLP中典型应用? 220 [填空题]
_________________________________
6.5 RNNs训练和传统ANN训练异同点? 220 [填空题] _________________________________
6.6常见的RNNs扩展和改进模型 221 [填空题] _________________________________
6.6.1 Simple RNNs(SRNs) 221 [填空题] _________________________________
6.6.2 Bidirectional RNNs 221 [填空题] _________________________________
6.6.3 Deep(Bidirectional) RNNs 222 [填空题] _________________________________
6.6.4 Echo State Networks（ESNs） 222 [填空题] _________________________________
6.6.5 Gated Recurrent Unit Recurrent Neural Networks 224 [填空题] _________________________________
6.6.6 LSTM Netwoorks 224 [填空题]
_________________________________
6.6.7 Clockwork RNNs(CW-RNNs) 225 [填空题] * _________________________________
7.1基于候选区域的目标检测器 228 [填空题] _________________________________
7.1.1滑动窗口检测器 228 [填空题] _________________________________
7.1.2选择性搜索 229 [填空题] _________________________________
7.1.3 R-CNN 230 [填空题]
_________________________________
7.1.4边界框回归器 230 [填空题] _________________________________
7.1.5 Fast R-CNN 231 [填空题] _________________________________
7.1.6 ROI 池化 233 [填空题]
_________________________________
7.1.7 Faster R-CNN 233 [填空题] _________________________________
7.1.8候选区域网络 234 [填空题] _________________________________
7.1.9 R-CNN 方法的性能 236 [填空题]
_________________________________
7.2 基于区域的全卷积神经网络（R-FCN） 237 [填空题] _________________________________
7.3 单次目标检测器 240 [填空题] _________________________________
7.3.1单次检测器 241 [填空题] _________________________________
7.3.2滑动窗口进行预测 241 [填空题] _________________________________
7.3.3 SSD 243 [填空题]
_________________________________
7.4 YOLO系列 244 [填空题]
_________________________________
7.4.1 YOLOv1介绍 244 [填空题] _________________________________
7.4.2 YOLOv1模型优缺点? 252 [填空题] _________________________________
7.4.3 YOLOv2 253 [填空题]
_________________________________
7.4.4 YOLOv2改进策略 254 [填空题] _________________________________
7.4.5 YOLOv2的训练 261 [填空题] _________________________________
7.4.6 YOLO9000 261 [填空题] _________________________________
7.4.7 YOLOv3 263 [填空题]
_________________________________
8.1 传统的基于CNN的分割方法缺点? 269 [填空题] _________________________________
8.1 FCN 269 [填空题]
_________________________________
8.1.1 FCN改变了什么? 269 [填空题] _________________________________
8.1.2 FCN网络结构? 270 [填空题] _________________________________
8.1.3全卷积网络举例? 271 [填空题]
_________________________________
8.1.4为什么CNN对像素级别的分类很难? 271 [填空题] _________________________________
8.1.5全连接层和卷积层如何相互转化? 272 [填空题] _________________________________
8.1.6 FCN的输入图片为什么可以是任意大小? 272 [填空题]
_________________________________
8.1.7把全连接层的权重W重塑成卷积层的滤波器有什么好处? 273 [填空题] _________________________________
8.1.8反卷积层理解 275 [填空题] _________________________________
8.1.9跳级(skip)结构 276 [填空题] _________________________________
8.1.10模型训练 277 [填空题]
_________________________________
8.1.11 FCN缺点 280 [填空题] _________________________________
8.2 U-Net 280 [填空题]
_________________________________
8.3 SegNet 282 [填空题]
_________________________________
8.4空洞卷积(Dilated Convolutions) 283 [填空题] _________________________________
8.4 RefineNet 285 [填空题]
_________________________________
8.5 PSPNet 286 [填空题]
_________________________________
8.6 DeepLab系列 288 [填空题] _________________________________
8.6.1 DeepLabv1 288 [填空题] _________________________________
8.6.2 DeepLabv2 289 [填空题] _________________________________
8.6.3 DeepLabv3 289 [填空题] _________________________________
8.6.4 DeepLabv3+ 290 [填空题] _________________________________
8.7 Mask-R-CNN 293 [填空题]
_________________________________
8.7.1 Mask-RCNN 的网络结构示意图 293 [填空题] _________________________________
8.7.2 RCNN行人检测框架 293 [填空题] _________________________________
8.7.3 Mask-RCNN 技术要点 294 [填空题]
_________________________________
8.8 CNN在基于弱监督学习的图像分割中的应用 295 [填空题] _________________________________
8.8.1 Scribble标记 295 [填空题] _________________________________
8.8.2 图像级别标记 297 [填空题]
_________________________________
8.8.3 DeepLab+bounding box+image-level labels 298 [填空题] _________________________________
8.8.4统一的框架 299 [填空题] * _________________________________
9.1强化学习的主要特点? 301 [填空题] _________________________________
9.2强化学习应用实例 302 [填空题]
_________________________________
9.3强化学习和监督式学习、非监督式学习的区别 303 [填空题] _________________________________
9.4 强化学习主要有哪些算法? 305 [填空题] _________________________________
9.5深度迁移强化学习算法 305 [填空题] _________________________________
9.6分层深度强化学习算法 306 [填空题] _________________________________
9.7深度记忆强化学习算法 306 [填空题] _________________________________
9.8 多智能体深度强化学习算法 307 [填空题] _________________________________
9.9深度强化学习算法小结 307 [填空题] * _________________________________
10.1 什么是迁移学习? 309 [填空题] _________________________________
10.2 什么是多任务学习? 309 [填空题] _________________________________
10.3 多任务学习有什么意义? 309 [填空题] _________________________________
10.4 什么是端到端的深度学习? 311 [填空题] _________________________________
10.5 端到端的深度学习举例? 311 [填空题] _________________________________
10.6 端到端的深度学习有什么挑战? 311 [填空题] _________________________________
10.7 端到端的深度学习优缺点? 312 [填空题] * _________________________________
13.1 CPU和GPU 的区别? 314 [填空题] _________________________________
13.2如何解决训练样本少的问题 315 [填空题] _________________________________
13.3 什么样的样本集不适合用深度学习? 315 [填空题] _________________________________
13.4 有没有可能找到比已知算法更好的算法? 316 [填空题] _________________________________
13.5 何为共线性, 跟过拟合有啥关联? 316 [填空题] _________________________________
13.6 广义线性模型是怎被应用在深度学习中? 316 [填空题] _________________________________
13.7 造成梯度消失的原因? 317 [填空题] _________________________________
13.8 权值初始化方法有哪些 317 [填空题]
_________________________________
13.9 启发式优化算法中，如何避免陷入局部最优解? 318 [填空题] _________________________________
13.10 凸优化中如何改进GD方法以防止陷入局部最优解 319 [填空题] _________________________________
13.11 常见的损失函数? 319 [填空题]
_________________________________
13.14 如何进行特征选择（feature selection）? 321 [填空题] _________________________________
13.14.1 如何考虑特征选择 321 [填空题]
_________________________________
13.14.2 特征选择方法分类 321 [填空题] _________________________________
13.14.3 特征选择目的 322 [填空题]
_________________________________
13.15 梯度消失/梯度爆炸原因，以及解决方法 322 [填空题] _________________________________
13.15.1 为什么要使用梯度更新规则? 322 [填空题] _________________________________
13.15.2 梯度消失、爆炸原因? 323 [填空题] _________________________________
13.15.3 梯度消失、爆炸的解决方案 324 [填空题] _________________________________
13.16 深度学习为什么不用二阶优化 325 [填空题] _________________________________
13.17 怎样优化你的深度学习系统? 326 [填空题] _________________________________
13.18为什么要设置单一数字评估指标? 326 [填空题]
_________________________________
13.19满足和优化指标（Satisficing and optimizing metrics） 327 [填空题] _________________________________
13.20 怎样划分训练/开发/测试集 328 [填空题] _________________________________
13.21如何划分开发/测试集大小 329 [填空题] _________________________________
13.22什么时候该改变开发/测试集和指标? 329 [填空题] _________________________________
13.23 设置评估指标的意义? 330 [填空题] _________________________________
13.24 什么是可避免偏差? 331 [填空题] _________________________________
13.25 什么是TOP5错误率? 331 [填空题] _________________________________
13.26 什么是人类水平错误率? 332 [填空题]
_________________________________
13.27 可避免偏差、几大错误率之间的关系? 332 [填空题] _________________________________
13.28 怎样选取可避免偏差及贝叶斯错误率? 332 [填空题] _________________________________
13.29 怎样减少方差? 333 [填空题]
_________________________________
13.30贝叶斯错误率的最佳估计 333 [填空题]
_________________________________
13.31举机器学习超过单个人类表现几个例子? 334 [填空题] _________________________________
13.32如何改善你的模型? 334 [填空题] _________________________________
13.33 理解误差分析 335 [填空题] _________________________________
13.34 为什么值得花时间查看错误标记数据? 336 [填空题] _________________________________
13.35 快速搭建初始系统的意义? 336 [填空题]
_________________________________
13.36 为什么要在不同的划分上训练及测试? 337 [填空题] _________________________________
13.37 如何解决数据不匹配问题? 338 [填空题] _________________________________
13.38 梯度检验注意事项? 340 [填空题] _________________________________
13.39什么是随机梯度下降? 341 [填空题] _________________________________
13.40什么是批量梯度下降? 341 [填空题] _________________________________
13.41什么是小批量梯度下降? 341 [填空题] _________________________________
13.42怎么配置mini-batch梯度下降 342 [填空题] _________________________________
13.43 局部最优的问题 343 [填空题] _________________________________
13.44提升算法性能思路 346 [填空题] * _________________________________
14.1 调试处理 358 [填空题]
_________________________________
14.2 有哪些超参数 359 [填空题]
_________________________________
14.3 如何选择调试值? 359 [填空题] _________________________________
14.4 为超参数选择合适的范围 359 [填空题] _________________________________
14.5 如何搜索超参数? 359 [填空题] * _________________________________
15.1 什么是正则化? 361 [填空题] _________________________________
15.2 正则化原理? 361 [填空题] _________________________________
15.3 为什么要正则化? 361 [填空题]
_________________________________
15.4 为什么正则化有利于预防过拟合? 361 [填空题] _________________________________
15.5 为什么正则化可以减少方差? 362 [填空题] _________________________________
15.6 L2正则化的理解? 362 [填空题] _________________________________
15.7 理解dropout 正则化 362 [填空题] _________________________________
15.8 有哪些dropout 正则化方法? 362 [填空题] _________________________________
15.8 如何实施dropout 正则化 363 [填空题] _________________________________
15.9 Python 实现dropout 正则化 363 [填空题] _________________________________
15.10 L2正则化和dropout 有什么不同? 363 [填空题] _________________________________
15.11 dropout有什么缺点? 363 [填空题] _________________________________
15.12 其他正则化方法? 364 [填空题] * _________________________________
因篇幅问题不能全部显示，请点此查看更多更全内容