tag:blogger.com,1999:blog-6141980.post5166747984398787240..comments2024-03-05T00:48:38.149-06:00Comments on Nuit Blanche: CS: Multiplicative Noise Effect on the Donoho-Tanner Phase Transition (Part 3)Igorhttp://www.blogger.com/profile/17474880327699002140noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-6141980.post-47435405650630727672011-04-08T06:02:08.268-05:002011-04-08T06:02:08.268-05:00JuanPi,
I mentioned what the axes were in the fir...JuanPi,<br /><br />I mentioned what the axes were in the first part of this entry (part 1). But I agree with you that the axis ought to be labeled in all the entries. I have not done a summary yet of this whole undertaking but basically I am reproducing the donoho-tanner transition for different solvers. As you can see this is work in progress but be reassured that in the summary version, I'll explain all these things so that a proper context can be infered from all these graphs.<br /><br />If you are coming from today's entry, the big surprise here is that the recovery of vectors using a least square solver requires the solution to be sparse in the case when there is noise in the measurement matrix. When there is no noise, we all know that the least square solver produces the best solution possible. The surprising finding is to find out that sparsity is an element in a case where people expect no such dependency.<br /><br /><br />Hope it helped,<br /><br />Cheers,<br /><br />Igor.<br />Igor.Igorhttps://www.blogger.com/profile/17474880327699002140noreply@blogger.comtag:blogger.com,1999:blog-6141980.post-83424181014337039632011-04-08T05:52:02.782-05:002011-04-08T05:52:02.782-05:00Hi,
your blog is interesting but sometimes fials t...Hi,<br />your blog is interesting but sometimes fials to comunicate to the nonCS expert. Maybe is not in your interest.<br /><br />In this case, I kind of follow the discussion, but the plots have no labels on their axis!(always put labels to your axis!) What are you plotting?JuanPi Carbajalhttps://www.blogger.com/profile/04710943376399349416noreply@blogger.com