Dear All
Do I solve this optimization problem in CVX?
w = 10^6; g = 1; b = 10^6; s = 10^(-15); cvx_begin variables p1 p2 minimize 1 - (p1 + p2) subject to w - (-b*inv_pos((p2*g + s)*log(2)))*rel_entr(p2*g + s, p1*g + p2*g + s) <= 0; p1 >= p2 cvx_end
inv_pos((p2*g + s)*rel_entr(p2*g + s, p1*g + p2*g + s) simplifies to log((p2*g + s)/(p1*g + p2*g + s)) which is neither convex nor concave. So I believe the first constraint, and therefore the optimization problem, are not convex,
inv_pos((p2*g + s)*rel_entr(p2*g + s, p1*g + p2*g + s)
log((p2*g + s)/(p1*g + p2*g + s))