A lot of people ask me to describe my thesis research. I used to give a complicated answer about using covariance matrices to perform a Karhunen-Loève transform to blah blah blah, but now I just say, “I clean cosmic data.” Today I created a graphic that illustrates the essence of what I’m trying to do.
Notice how both the signal and noise have “structure”? My research attempts to uncover those structures and use them to eliminate the noise. The problem is that taking away noise also takes away signal. So we need a way to “fill in the gaps.”
Now imagine doing this not for images, but for MASSIVE data sets…and you don’t get to know what the letters are beforehand. Solving this problem to high accuracy is a challenge.