Consider $f$ to be a convex multivariate function from $\mathbb{R}^N$ to $\mathbb{R}$, not differentiable (but, by definition, left- and right-differentiable).

Does some kind of "directional" 2nd-order approximation formula exist, for example where $$ f(x+ty) = f(x)+<v,ty>+\frac{1}{2}<Ay,y> $$

for $(x,y)\in\mathbb{R}^N$ and a scalar $t>0$ and $v$ is a subgradient of $f$ at $x$ and $A$ is a matrix? When $f$ is convex not differentiable, I struggle to justify that.

Thanks for your help.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.