Tuesday, October 11, 2011

"What's an algorithm ?"

Meanwhile in the real world people gasp a huge Wow when confronted with some simple deblurring;  [the video below is now private, it showed s person from Adobe/Photoshop featuring a deblur button, a feature potentially inserted in future Adobe products. As the presenter went on showing some examples of the "deblur" button, the crowd kept on gasping with wows, underlining the clarity of the deblured pictures. another person on stage playing the role of the interviewer then asked how this magic could be enabled. The presenter told him it's a new algorithm. The stage person then went on to incredulously ask "what's an algorithm ?". Hence the title of this entry ]

(via PetaPixel)

In a somewhat totally different direction and within the business of science publishing here is a potentially egregious behavior that I will call out on this blog. Why egregious ? Because we most probably have had this type of behavior before. Let us recall that currently one of the best solvers out there in terms of both speed and how far it goes beyond the Donoho-Tanner phase transition (SL0) received a disgraceful rejection letter from the IEEE Signal Processing letters in freaking 2006. I personally asked the authors to kindly put it up on their website (since the claim was so astounding) and they did, first on this blog then on the website. To put this in perspective, a near two full years before the rest of the community gets around to having a fast and similarly efficient solver, the IEEE Signal Processing Letters decided through its own process that this paper was not worthy.  Maybe the process is not worthy of real science. I know ... I know... you are going to say: "What could possibly replace peer-review ?", "How are we going to know what paper is important, we can't even read the papers featured on Nuit Blanche ?". I am not sure but a process that keeps the players honest would sure beat this one.

I also noticed that a large section of retracted papers featured on Retraction Watch could have been caught automatically through the use of a good image comparison algorithm. If the publication were open, I am sure some enterprising folks could take a stab at this one.

Liked this entry ? subscribe to Nuit Blanche's feed, there's more where that came from. You can also subscribe to Nuit Blanche by Email, explore the Big Picture in Compressive Sensing or the Matrix Factorization Jungle and join the conversations on compressive sensing, advanced matrix factorization and calibration issues on Linkedin.

No comments: