Week 2

Welcome back! This week went by fast and in terms of our project, I think our group is making good progress - and it's only week 2! Last week, we were given code in Matlab and since our group decided to work in Python, we had to translate all the code into Python. So this week we had the code and started to solve the inverse problem with data from sea surface temperatures (SST). Our goal was to reconstruct the true image of SST as best as we could with the given data (i.e. collected SST data from sensors at certain locations around the world). Our first attempt at solving this problem gave us a relative error of about 44% (not great, so we tried to bring this error down). After spending a couple of days looking at the data and plotting other aspects of it, we noticed that if we scaled the posterior covariance, the error went down by about 5-10%. This was because scaling accounted for the added error from the truncated singular value decomposition. While I took an intro class to Python, I'm still learning how to code, especially with big datasets, but I think our plots came out pretty nice!

The reconstruction plots of the SST data and its standard deviations.

On Wednesday, all of us had to attend a seminar on technical writing. We learned what goes into a research paper, how to write an introduction, and the differences between math, stats, and bio papers. Then the next day, we met with our advisor and showed him the work we've done so far with the plots. He told us we're making good progress with the inverse problem and suggested exploring another dataset (Cylinder Flow With Von Karman Vortex Street Data). Currently, we're focusing on solving the inverse problem with a fixed given number of sensors and trying to replicate and understand the results of the first paper we read. Once we have a good understanding of this, we are going to move on to explore how to optimally choose these sensor locations (which is from the second paper we read), and then add in the new part of the multi-fidelity aspect (how to optimally choose and place cheap, high-noise sensors vs expensive, low-noise sensors) of the problem - which is the main focus of this project. Anyway, on Friday we explored the new dataset and plotted its reconstruction and then began deriving the equations for the multi-fidelity problem! We also attended another seminar in the afternoon on how to present yourself professionally, and then had our weekly tea/coffee with the whole group. This week my roommate's project group had to pick the social activity, and they decided on fishbowl. I've never heard of the game before but it was essentially charades with a twist. After that, we went back to our apartments and some of us met up again in the common room in our apartment building for movie night! We decided to do movie nights every Friday, and this week the movie was "Scream". We watched the first one that was from the 90s, and it was good! On Saturday, some of us wanted to do another hike at Umstead State Park, so we carpooled there. It was definitely hotter than last week, but the trail was nice. After that, we went to the international food festival in downtown Raleigh. It is a big event that is hosted every year, and there is food and music from all over the world! It was a nice day but then a thunderstorm rolled in and we all got soaked trying to get back to our cars. It was quite the experience! Luckily most of us already bought food and walked around, so after the storm ended, we drove back to our apartments. It was definitely a fun evening!

International Food Festival!

Thanks for reading, see you next time!

Gaby