Uploaded February 2021 | Updated November 2026, 3 hours ago
NOTE: To simplify the discussion of KKT multipliers, we assume that the constraint is active. To work this assumption into the problem, you need a condition called complimentary slackness, which isn't discussed in the video.
To work out the kernel trick for SVMs, we need to know how to apply the method of Lagrange multipliers to a constrained optimization problem. We discuss both the classical case, with equality constraints, leading to the method of Lagrange multipliers and the modern case with inequality constraints, leading to KKT multipliers.
slides: mlvu.github.io/lectures/32.LinearModels2.annotated.pdf
lecturer: Peter Bloem
NOTE: To simplify the discussion of KKT multipliers, we assume that the constraint is active. To work this assumption into the problem, you need a condition called complimentary slackness, which isn't discussed in the video.
To work out the kernel trick for SVMs, we need to know how to apply the method of Lagrange multipliers to a constrained optimization problem. We discuss both the classical case, with equality constraints, leading to the method of Lagrange multipliers and the modern case with inequality constraints, leading to KKT multipliers.
slides: mlvu.github.io/lectures/32.LinearModels2.annotated.pdf
lecturer: Peter Bloem










