II
6.2 假说表示 .................................................................................................................. 98
6.3 判定边界 ................................................................................................................ 100
6.4 代价函数 ................................................................................................................ 102
6.5 简化的成本函数和梯度下降 ................................................................................ 105
6.6 高级优化 ................................................................................................................ 108
6.7 多类别分类:一对多 ............................................................................................ 112
七、正则化(Regularization) ................................................................................................. 115
7.1 过拟合的问题 ........................................................................................................ 115
7.2 代价函数 ................................................................................................................ 117
7.3 正则化线性回归 .................................................................................................... 119
7.4 正则化的逻辑回归模型 ........................................................................................ 120
第 4 周 .......................................................................................................................................... 122
第八、神经网络:表述(Neural Networks: Representation)............................................... 122
8.1 非线性假设 ............................................................................................................ 122
8.2 神经元和大脑 ........................................................................................................ 124
8.3 模型表示 1 ............................................................................................................. 128
8.4 模型表示 2 ............................................................................................................. 132
8.5 特征和直观理解 1 ................................................................................................. 134
8.6 样本和直观理解 II ................................................................................................. 136
8.7 多类分类 ................................................................................................................ 138
第 5 周 .......................................................................................................................................... 139
九、神经网络的学习(Neural Networks: Learning) ............................................................. 139
9.1 代价函数 ................................................................................................................ 139
9.2 反向传播算法 ........................................................................................................ 141
9.3 反向传播算法的直观理解 .................................................................................... 144
9.4 实现注意:展开参数 ............................................................................................ 147
9.5 梯度检验 ................................................................................................................ 148
9.6 随机初始化 ............................................................................................................ 150
9.7 综合起来 ................................................................................................................ 151
9.8 自主驾驶 ................................................................................................................ 152
第 6 周 .......................................................................................................................................... 155
十、应用机器学习的建议(Advice for Applying Machine Learning) ................................... 155
10.1 决定下一步做什么 .............................................................................................. 155
10.2 评估一个假设 ...................................................................................................... 158
10.3 模型选择和交叉验证集 ...................................................................................... 160
10.4 诊断偏差和方差 .................................................................................................. 162
10.5 归一化和偏差/方差 ............................................................................................ 164
10.6 学习曲线 .............................................................................................................. 166
10.7 决定下一步做什么 .............................................................................................. 168
十一、机器学习系统的设计(Machine Learning System Design) ....................................... 170
11.1 首先要做什么 ...................................................................................................... 170
11.2 误差分析 .............................................................................................................. 171
11.3 类偏斜的误差度量 .............................................................................................. 174
11.4 查全率和查准率之间的权衡 .............................................................................. 175