Three things

[Kevin Waugh]

(a) In lecture yesterday I miswrote the divergent step size rule.  My condition was sufficient, but not necessary (as pointed out by Aaditya as 1/sqrt(k) did not satisfy it).  The proper condition is

t_k > 0, t_k -> 0, sum t_i -> infinity

(b) In homework 5, 1b) Hint: if you look close enough, the y should look an awful lot like -c.

(c) In homework 5, 1) throughout the problem, you may make any assumptions necessary to assume solutions to the problems exist (i.e., appropriate inverses, rank constraints on constraint matrices etc.).  You don't have to state these assumptions.