Oops!

So, my FSM corrections were wrong. Fortunately, the CS faculty member on my committee was able to straighten me out today. I had tried to divide the vector of inner products of the sensitivities and errors by the norm of the sensitivity matrix. Apparently, this is only allowed if you are dealing with a scalar. Oops.

But, after minor code revisions, I was able to compute the corrections correctly. The funny thing is that the corrections are better when you correct them correctly. Imagine that.