After an absolutely amazing week of vacationing in Reston, Virginia, I finally came back to work today ready to catch up on what I missed and help the team again. My morning was spent for the most part trying to catch up by going through missed emails, reading the paper for our reading group, looking over new code that was pushed, and going to CSE to get Leonardo (my beloved work laptop) back. Before I knew it, 10:30 AM had rolled around and I had to go meet with the TCAT group back in CSE. Our meeting this week consisted mainly of quick updates on how each of the projects are going, and Anat updated us about how there will be high-schoolers coming in this week to help with the access map project. Also next week the goal is to try out the battery interrupter circuit project to make sure we can assemble it easily and efficiently. After our meeting, I had to quickly head back to the eScience center so I could be on time for our 11 AM reading group discussion meeting. Honestly, I don't think I got much out of this week's reading group discussion. We talked a lot about census data and the distinction between organic and designed data, and how designed data can be better in some ways because it better represents the population. It was a bit interesting at first, but talking about it for an hour seemed to be a bit of a stretch for me.
After lunch I familiarized myself more with the slides for our big presentation on Wednesday. The presentation is supposed to be 8 minutes long, allowing two minutes for questions at the end and it will be presented in front of the eScience Steering Committee. Also, Valentina told us that there might be a reporter coming, and that she believes he is supposed to talk to our group after the presentations. That's pretty exciting. So, Kristen had updated the slide that explained our algorithm, but then she had to leave, so I looked over the changes she made and then I made some further changes myself. I also timed myself just speaking the notes at the bottom of every slide and it came to eight and a half minutes. That worried me a little because I think it's highly likely we'll go over our time, and I think that a lot of people will also have questions since our problem can be a bit confusing at first.
Additionally, in the afternoon we all took some time to talk about the algorithm again, specifically how we were going to do the pickup and drop-off insertions and how they will deal with the time windows. Unsurprisingly, we ended up running into our constant struggle between having a comprehensive response and keeping our program from being too slow and computationally expensive. At the present time, we have decided to take only the best solution from the pickup insertion, and use that to find drop offs, and if that yields us no results too often we can raise the amount of insertion points that we will go off of, using maybe the three best options or five best options. The reason we don't want to do this right away is because each different insertion point requires more and more calls to the osm routing function which is slow and computationally expensive, and so we are trying to limit the number of calls we make to it.
After lunch I familiarized myself more with the slides for our big presentation on Wednesday. The presentation is supposed to be 8 minutes long, allowing two minutes for questions at the end and it will be presented in front of the eScience Steering Committee. Also, Valentina told us that there might be a reporter coming, and that she believes he is supposed to talk to our group after the presentations. That's pretty exciting. So, Kristen had updated the slide that explained our algorithm, but then she had to leave, so I looked over the changes she made and then I made some further changes myself. I also timed myself just speaking the notes at the bottom of every slide and it came to eight and a half minutes. That worried me a little because I think it's highly likely we'll go over our time, and I think that a lot of people will also have questions since our problem can be a bit confusing at first.
Additionally, in the afternoon we all took some time to talk about the algorithm again, specifically how we were going to do the pickup and drop-off insertions and how they will deal with the time windows. Unsurprisingly, we ended up running into our constant struggle between having a comprehensive response and keeping our program from being too slow and computationally expensive. At the present time, we have decided to take only the best solution from the pickup insertion, and use that to find drop offs, and if that yields us no results too often we can raise the amount of insertion points that we will go off of, using maybe the three best options or five best options. The reason we don't want to do this right away is because each different insertion point requires more and more calls to the osm routing function which is slow and computationally expensive, and so we are trying to limit the number of calls we make to it.