Tuesday, April 14, 2009

What is NOT Compressed Sensing?

Will someone (perhaps the phenomenally committed Igor Carron with his compressed sensing blog or the Rice folks) explain to the world what is NOT compressed sensing. It is NOT
  • Any sparse representation with small number of nonzeros.
  • Any dimensionality reduction by random projections.
  • Any operation that replaces L_0 optimization by L_1.
  • Any heavy hitters algorithm.
I do not want to listen to so many machine learning, signal processing, data mining, statistics, image processing researchers tell me that they do compressed sensing. Or may be I should throw up my hands and let the world embrace Compressed Sensing, after all they are sensing something, and it is compressed.


Anonymous Anonymous said...

You should explain why it matters that things are mislabeled. Do Compressed Sensing techniques apply to the things you consider *not* to be Compressed Sensing? If so, what is the harm in mislabeling? Is the harm that techniques are incorrectly applied?

6:10 AM  
Blogger Igor said...



The short of it is: Throw up your arms :-)


2:59 AM  
Blogger S said...

Hi Igor:
Thanks for the note and the post (information theory timeline is cool!).

Hi Anon:
If compressed sensing techniques applies to things, that is great. But if someone takes whatever they do and call it compressed sensing, because it sorta sounds like the English words "compressed" and "sensing", it feels like some harm has been done, but may be it is just in a cosmic sense.

My arms are up.

8:58 AM  
Anonymous Anonymous said...

1) it's bad because it had, originally, a precise meaning: a particular loss function and a particular algorithm for minimizing that loss. now it has been robbed of its meaning.

2) you can help it retain its meaning by introducing terms like "strong and weak CS" or "strict and loose" CS or something to correspond to different loss functions or different algos for minimizing them.

3) i'm sad to hear CS has joined nanotechnology, systems biology, and multiscale as words one says to excite program officers rather than precisely-formed optimization tasks.

11:54 AM  
Blogger Igor said...


I don't really think that the current situation is as bad as what you seem to imply in point 3). In the end, what is important is whether CS provides a means of actually doing things that we could not do otherwise. On the hardware side, this is still being hammered with some promising leads.

FWIW, I am trying to compile a list here:



12:36 PM  

Post a Comment

<< Home