2 Comments
User's avatar
rif a saurous's avatar

Nit: "If potential decreases at every time step but is always positive, it eventually has to get to zero." You need a little more, right? Geometric decay, which shows up as the epsilon later when you math it out?

Ben Recht's avatar

Right, you don't need geometric decay, but you need some extra assumption. In the reading list, the notes by Boyd and Lessard give a few examples of the additional assumptions needed for full rigorous proofs.