Projects Flashcards
NASA Capstone Overview
This project was part of our Capstone during my masters. The overlying goal was the improve estimates of PM2.5, which is essentially refers to fine inhalable air particles.
Currently the way that PM2.5 is measured is through ground sensors throughout the country. These sensors are very accurate but obviously expensive and not very scalable. Now basically a bunch of very smart PhDs learned that they could accurately retrieve aerosol estimates including PM2.5 using polarimeter data. There’s something called the PACE-MAPP algorithm that allows you to do this. But the problem is this model is very computationally intensive and so the idea of our project is to replace the inefficient portion of this algorithm with a neural network.
And we can go into more detail here in a bit but long story short we were able to speed up the heavy portion of the algorithm by over 1000x while maintaining a very high level of accuracy.
