An explanation of the meaning of duality. It's all in the gradients!
Errata:
13:12 I said that (c1,c2) is v1 times the first gradient plus v1 times the second gradient. I should have said, v1 times the first gradient plus v2 times the second gradient.
16.51 I said that x is optimal (maximizes cx over all feasible x'). But the inequality shows that cx' is less than or equal to vb. I should have emphasized that vb = cx, so this is the same as saying that cx' is less than or equal to cx.
17:45 The same thing here: the inequality shows that v'b is greater than or equal to cx. But since cx =vb, that means that v'b is greater than or equal to vb, so v'b is minimized when v' = v (the dual solution).
