II
6.2 假说表示 .................................................................................................................. 53
6.3 判定边界 .................................................................................................................. 55
6.4 代价函数 .................................................................................................................. 57
6.5 简化的成本函数和梯度下降 .................................................................................. 60
6.6 高级优化 .................................................................................................................. 61
6.7 多类分类:一个对所有 .......................................................................................... 62
七、正则化(Regularization) ................................................................................................... 63
7.1 过拟合的问题 .......................................................................................................... 63
7.2 代价函数 .................................................................................................................. 65
7.3 正则化线性回归 ...................................................................................................... 67
7.4 正则化的逻辑回归模型 .......................................................................................... 68
第 4 周 ............................................................................................................................................ 69
第八、神经网络:表述(Neural Networks: Representation)................................................. 69
8.1 非线性假设 .............................................................................................................. 69
8.2 神经元和大脑 .......................................................................................................... 71
8.3 模型表示 1 ............................................................................................................... 75
8.4 模型表示 2 ............................................................................................................... 79
8.5 特征和直观理解 1 ................................................................................................... 81
8.6 样本和直观理解 II ................................................................................................... 83
8.7 多类分类 .................................................................................................................. 85
第 5 周 ............................................................................................................................................ 86
九、神经网络的学习(Neural Networks: Learning) ............................................................... 86
9.1 代价函数 .................................................................................................................. 86
9.2 反向传播算法 .......................................................................................................... 88
9.3 反向传播算法的直观理解 ...................................................................................... 91
9.4 实现注意:展开参数 .............................................................................................. 94
9.5 梯度检验 .................................................................................................................. 95
9.6 随机初始化 .............................................................................................................. 97
9.7 综合起来 .................................................................................................................. 98
9.8 自主驾驶 .................................................................................................................. 99
第 6 周 .......................................................................................................................................... 102
十、应用机器学习的建议(Advice for Applying Machine Learning) ................................... 102
10.1 决定下一步做什么 .............................................................................................. 102
10.2 评估一个假设 ...................................................................................................... 105
10.3 模型选择和交叉验证集 ...................................................................................... 107
10.4 诊断偏差和方差 .................................................................................................. 109
10.5 归一化和偏差/方差 ............................................................................................ 111
10.6 学习曲线 .............................................................................................................. 113
10.7 决定下一步做什么 .............................................................................................. 115
十一、机器学习系统的设计(Machine Learning System Design) ....................................... 117
11.1 首先要做什么 ...................................................................................................... 117
11.2 误差分析 .............................................................................................................. 118
11.3 类偏斜的误差度量 .............................................................................................. 121
11.4 查全率和查准率之间的权衡 .............................................................................. 122