Introduction to Fenchel Duality for Machine Learning - middle angle
Introduction to Fenchel Duality for Machine Learning - middle angle
Introduction to Fenchel Duality for Machine Learning - middle angle
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
The Differentiable Case<br />
Suppose f and g are both convex and differentiable. Then the<br />
complementarity conditions become<br />
which combine in<strong>to</strong><br />
z = ∇f(y)<br />
−z = ∇g(y),<br />
∇f(y) + ∇g(y) = 0,<br />
the well-known optimality condition <strong>for</strong> f + g.<br />
R. Rifkin <strong>Fenchel</strong> <strong>Duality</strong> I<br />
honda-logo