I like your acronym Igor, and the self-deprecating humor, but it won't sell an idea! :)
I think your algorithm is not the same. With the ensemble of solvers, the measurements are already made. We just choose the reconstruction giving the lowest error. We don't make any assumption that the set of signals recovered by OMP includes the set of signals recovered by BP, for instance. I have found the two sets overlap, but one is not a subset.
The idea of DUMBEST is to use only the measurements you have with a series of different solvers and see if some sort of correlation can tell you a solution (without the oracle of the DT computation where you know the solution). So in effect, I really think it is of the same vein. Let me expand on this a little further in an entry.
3 comments:
I like your acronym Igor, and the self-deprecating humor, but it won't sell an idea! :)
I think your algorithm is not the same. With the ensemble of solvers, the measurements are already made. We just choose the reconstruction giving the lowest error. We don't make any assumption that the set of signals recovered by OMP includes the set of signals recovered by BP, for instance. I have found the two sets overlap, but one is not a subset.
Bob,
The idea of DUMBEST is to use only the measurements you have with a series of different solvers and see if some sort of correlation can tell you a solution (without the oracle of the DT computation where you know the solution). So in effect, I really think it is of the same vein. Let me expand on this a little further in an entry.
Cheers,
Igor.
Post a Comment