Minimizing a linear function over the probability simplex

932 Views Asked by At

I am trying to derive the solution of the problem below but am having trouble

enter image description here

This is my solution

$L = c^Tx + v(1^Tx - 1) - \lambda^Tx \\ \ \ \ = (c + v1 - \lambda)^Tx - v $

The dual is thus

$\min v \\ st. v1 = \lambda - c \\ \lambda \ge 0$

I dont know how to arrive at theirs. I also don't understand my solution. I have seen the solution from boyd's book on convex optimization but I want to do it through the dual for practice.