Support-Vector Networks

Introduced the soft-margin hyperplane formulation allowing classification errors via slack variables in Support Vector Machines.

Core algorithm implemented in LibSVM and scikit-learn for high-accuracy classification on small, high-dimensional datasets.

Assumptions

• Non-separable training data can be handled by balancing margin width against training error penalty C

Limitations

• Selecting appropriate kernel and penalty C requires grid search; training O(n^2) to O(n^3) memory/compute