After years of studying, I have started feeling pretty confident in my machine learning abilities. But my collaboration with SRON brings me face to face with my lack of experience with satellite imagery. Let’s wind back to your student days with me, and dive into the world of the pure joy of learning new things!
In my research, I try to find machine learning solutions that are general enough so they can solve different tasks. Though I collaborate with SRON to improve the detection of methane plumes, I do not want to come up with a new approach that is highly specific to this. I want people to be able to use my approach to do other things as well. Deep learning is very suited to this problem, because you can automatically extract features from your data. This means that you are not constrained to manually calculated features that may only be relevant to methane, but not to another gas, like carbon monoxide.
Now, I work with a dataset of methane plumes, created by a PhD student at SRON. But I will need to find other datasets to test how well my methods work with other problems. Luckily, another PhD student has a file with carbon monoxide plume locations for me: this is a nice test case, because it is pretty related to the detection of methane plumes, so my approach should work here, too. Unluckily, I need to find and process the corresponding satellite imagery myself: downloading and using satellite imagery is not exactly standard in a computer scientist’s toolbox. While I feel confident in my abilities to develop and use machine learning models, I feel totally lost with this task: I have no clue where to start. So, back to school, I go!
It gives me a big confidence boost that I can now easily solve problems I couldn’t as a master’s student. I remember how impossible it felt to get any neural network to train properly. A few weeks ago, I fixed the training of two different networks in a matter of minutes. It makes me feel like an expert. Moreover, it’s nice to be able to help people to solve these problems, especially researchers from different fields like at SRON by explaining machine learning concepts or sharing resources and useful python packages.
Being the person I am, this teacher role stresses me out sometimes. Because I’m good at what I do, I feel like I have to be good all the time and be a good example to people who want to learn more about coding and machine learning. I’m a bit jealous of the people I introduce to the wonders of machine learning: I yearn for that responsibility-free experience of simply learning.
Fortunately, at SRON I’m not only the local computer science expert, who’s there to preach the word of good coding practice and machine learning. I also sit on the other side of the bench: the pupil, the Earth Sciences noob, who struggles with navigating and using the data. From an ML perspective, I’m used to being fed more or less directly usable data, but the Earth Sciences perspective is all about that data. Feeling this helpless with the seemingly simple task of constructing a dataset throws me back to being a student again. And guess what – that’s fun!!
The noob status means I can forget about some of my teaching responsibility and just start playing around. It gives me a licence to make mistakes and enjoy the process. It does take some effort to let go, however. Like this weekend, when I went on a double date in a pool cafe. I was the absolute worst player (tbh, this is true for any game /sport involving a ball). This was anguish-inducing in the beginning, but quickly I realised that I can just enjoy, have fun and learn from my betters.
Hot take: learning new stuff is fun, and should stay fun. I want to be like my granddad. He regretted never having learned how to draw earlier in life, so he bought a beginner’s drawing kit at the supermarket. Like him, I don’t want to grow “old” and set in my ways too soon. I admire this shameless, playful, even hungry attitude towards life. It’s like Picasso said – “We are all born children – the trick is remaining one.”
Aside from the challenges, interdisciplinary research forces me out of my familiar ivory tower. While it does feel like reinventing the wheel sometimes, it feels great to experience how much I still have to learn. It’s so easy to become too serious as you become more experienced. So, I hope there will be plenty more unfamiliar datasets and problems in my future. Even more, I hope I can become young like my granddad.
February 19, 2023
EssaysThe impostor syndrome - every PhD student gets it at some point, or so I’m told. For me, it’s hard to disentangle from my own perfectionism: I’m tempted to think whatever I do is not good enough, regardless of how good I think I am at something compared to other people.
May 3, 2023
EssaysFirst of all: I’m a gym addict. The type of gym that looks like a big garage, has a lot of free weights and free live entertainment in the form of elite athletes (Sportja Leiden, if you’re interested ;) ). When we have to run during a class, that usually raises collective moans.
