ML Notes
Typical training parameters for different problem types
| Problem type | Loss function | Metrics | Final Activation Function |
|---|---|---|---|
| Binary classification | binary_crossentropy | accuracy | sigmoid |
| Multiclass classification | categorical_crossentropy | accuracy | softmax |
| Scalar regression | mse | mae | None |
Links
- Andrew Ng's Stanford ML Course on YouTube
- https://course.fast.ai/ Options for cloud computing setup for ML. Parent site has lots of other great resources.
- https://explained.ai/ Tutorials on some fundamental ML/DL topics, such as this inroduction to Matrix math for DL.
- http://tensorlab.cms.caltech.edu/users/anima/cms165-2020.html Caltech course with links to talks on YouTube and slides