Theory of Intelligence
MACHINE LEARNING•Classic Algorithms•Support Vector Machines•1.5 Under the Hood

Introduction

How a linear SVM makes predictions with its decision function, how the hinge loss measures margin violations, and why the dual problem enables kernels.

Introduction

You can use SVMs perfectly well without knowing their internals. But a look under the hood answers some natural questions:

  • What exactly is the "street", in numbers?
  • How does the soft margin decide how much to penalise a violation?
  • Why do kernels work only with some SVM solvers?

We'll keep the maths as light as possible and focus on the ideas.

Discuss theoretical concepts, mathematical formulas, and architectural foundations.

Supports inline code with `backticks`
No comments yet. Be the first to start the discussion!