Let
be a convex set in a real vector
space. A function
is nonconvex if it is not a convex
function; equivalently, there are
and
such that
A nonconvex function can have multiple local minima, so a local optimization method need not find a global minimum.