Machine Learning Fundamentals · 免费预览
本课程未涵盖的内容
本课程各讲所构建的流水线以可微的目标函数对定长输入进行评分，而处理序列、图像和生成任务的架构是在同一循环上加以扩展，而非将其取而代之。
1. The scaffold, restated as components
The scaffold, restated as components
2. Extensions, each touching one component
Extensions, each touching one component
3. Where the shape assumptions break first
Where the shape assumptions break first
4. What to check when a run goes wrong
What to check when a run goes wrong
5. Which component a new architecture changes
A practitioner replaces a fully connected layer with a convolutional layer in an image classifier. Which parts of the training procedure change?