Galois Summer School!

1st Galois Summer School!

• Yes! I was 1 of 35 persons chosen to attend the Galois Summer School this year on Trustworthy Machine Learning. It was awesome!
• I also met a bunch of people!
• Everyone was super nice!

Start

• I actually got there 15 minutes “late” because I was wrapping up a meeting, but as it turns out, I was just in time for the first slide!
• We started with a plenary, talking about Galois and what they do, and details for the day. They then gave us our first scenario and problem to work on in a group of about 4 (I was in group 2) with an assigned mentor.
• We were tasked with talking about our process and giving a report. It’s all recorded, so it may very be out in the magical world soon!
• Don’t dislike the video just ‘cause I’m in it; I will find you! :P
• Our first scenario involved trying to figure out based on a hospital admittances dataset who was likely to be readmitted. We decided on the lecture that we would use an SVM model. We also spoke about the assumptions, and the dataset, which had some unique attributes that were a bit tricky!

Break and 2nd Half

• After our presentation, we had lunch, and then headed into our second plenary, we heard a bit of the second talk, which included ideas we should think about, such as boosting, bagging and random forest models, as well as cross-validation. Our second dataset involved earthquake data.
• At this point we merged with group 6, and were thereafter referred to as “the super group”.
• What we found is that we weren’t sure this was even something you might necessarily want to use machine learning for at all. So we spoke about that.
• Some of our ideas involved agent modelling and clustering, or latent models, as well as other bayesian approaches. We finally spoke about how our evaluation and what we learned.

It was a swell day!

• We had a really upbeat mentor who really was quite enthusiastic and helpful, and smart. And everyone treated us quite well. Everyone had the opportunity to be heard.
• It is to be noted that this was their first of such an event, and so you should definitely look forward to more! I’d highly recommend it; it was absolutely worth attending and very fun, and the goal was not to build a complete model, but rather, to think about and look at creative ways of solving problems, and to talk about the constraints of dealing with each problem. So that was pretty cool.

That’s about all I have for now…so I guess that’s it!