Glossary · Phase 3 · Machine learning
Kernel trick
Replacing dot products with a similarity function so a linear method draws curved boundaries without building new features.
Try it now
The order of a real scikit-learn pipeline: data, a scaler and the model joined, then fit and score.
Learn it properly
Support vector machines · The widest margin, support vectors, the hinge loss and C, kernels for curved boundaries, and when an SVM fits. About 30 minutes.
Other terms from the same lesson
- hinge loss: max(0, 1 - y·score): zero for points safely on their side, growing for points inside the margin or on the wrong side.
- support vector machine: A classifier that picks the boundary with the widest margin to the nearest training points of each class.
Never trained an AI?
Teach a small one to write, right in your browser. It takes about 20 seconds and needs no account.