Skip to content

Support Vector Machines

Authors: Boser, Guyon, Vapnik Year: 1992-1995 Citations: 50,000+

Summary

Powerful supervised learning for classification/regression. Maximum margin hyperplane with kernel trick for non-linear boundaries. Dominated ML before deep learning.

Key Concepts

  • Maximum Margin: Optimal hyperplane maximizes distance to nearest examples
  • Support Vectors: Examples closest to decision boundary
  • Kernel Trick: Implicit non-linear mapping via kernels
  • Soft Margins: Handle non-separable data via slack variables
  • Dual Optimization: Quadratic programming formulation

Impact

  • 50,000+ citations - dominated 1990s-2000s ML
  • Best performance on many benchmarks
  • Strong theoretical generalization bounds
  • Still widely used for small-to-medium datasets
  • Foundation for kernel methods
  • Kernel Methods (Scholkopf et al., 1997)
  • Structural SVMs (Tsochantaridis et al., 2004)