Representation
How can machines represent increasingly complex patterns?
Move beyond straight lines by changing the features, the model shape, or the space where patterns are expressed.
1. Feature Engineering
Improve performance by changing the representation before changing the model.
2. Decision Trees
Replace one global equation with a hierarchy of local decisions.
3. Random Forests
Reduce tree variance by averaging many trees
4. Gradient Boosting
Correct errors one weak model at a time
5. Support Vector Machines
Find boundaries with maximum margin
6. Kernel Methods
Make non-linear patterns look linear
7. Representation Learning
Learn useful features instead of handcrafting them
8. Embeddings
Represent objects as meaningful vectors
9. Principal Component Analysis
Compress data while preserving structure
10. Clustering
Group examples without labels