AcademicalNerd @academicalnerd@qoto.org
People are more important than wires.
Joined Jan 2021
Working with supercomputers is a pain in the ass. It's a miracle these things even boot up with so many devices to keep track of, but using software gets far more weird. Half the operations work on one version and another half works on the other, but only if the number of used processors is even... Maybe it's ompi shenanigans, who knows. Weird.
Solving dependencies for conda packages used in computational chemistry is spectacularly annoying. But while I'm at it -- why not write something here.
It's been three months of work with machine learning, and over four full years of theoretical modelling overall. I learned an important lesson: the values your model predicts are not important. It is relatively easy to make a prediction using numpy, scipy and a bunch of polynomials.
It is far more important to know the applicability domain of the model, its error and trust intervals of obtained parameters. Clean the dataset, reduce the number of parameters, write out the necessary equations and try to reduce the error using first-principle approach. Modern science studies deviations and reproducibility, and rarely values themselves. Which is a good thing.
It's good, but I'm looking for something more forum-like.
Conferences now became quite close-minded and elitist: the same people, year after year, polished presentations, barely any input at all. There are a few large ones every year, but they are expensive and quite limited number of people are allowed in. Which raises the question: does anyone know a platform or a website where one could post unpolished results and discuss them with other in hopes of feedback, corrections or even collaboration?
There are few ways to do that, the trick is to have all the structures in the dataset have a representation of the same dimensionality, otherwise the neural network won't work. I use slightly modified SOAP descriptor from dscribe package (smooth overlap of atomic positions). You can skim through the docs if interested: https://singroup.github.io/dscribe/1.0.x/tutorials/descriptors/soap.html
I'm building a neural net that predicts stability of solid phases. The details are not quite settled yet, there's a lot of trial going on. Generally it's a classifier that guesses if energy above the hull is greater than zero or is equal to zero.
Convex hull is a thermodynamic concept: set of compounds that are thermodynamically stable, and their unstable polymorphs are higher in energy (hence E above the hull).
@rastinza
I'm not the first one to do this, luckily. Materialsproject contains enough data for my purposes and there are other databases to support it.
I've been working with a bunch of ML and quantum chemistry lately. At this point the hardest part is to actually come up with the problem (i.e. translate chemistry into reasonable mathematical expressions) and then make all the servers and packages work. The code and NN architecture themselves are like 10-20% of the work.
Now, a more personal note. For now I stay: to finish the education, to help the family. In some sense to see what happens next from the inside and maybe to tell the world. I will not be drafted in the near future and I think I will come up with a bailout in the next few months just to be on the safe side.
What has been happening over the last 6 months is awful and wrong and I hope it won't last for long. The consequences will be dire for everyone. God help us, even though I do not think there is one.
An important thing most people do not consider now is the economical impact. At least 300k working men are drafted, possibly more. The state feeds them, pays them a salary (which is not that small), pays substantial compensations in case of injury or death, while the economy is deprived of the 300k of perfectly good workers. It is unlikely to cause the collapse, but the consequences will hit after a while, I'm sure.
Well, that did not go well, eh. As some of the subscribers probably know, I'm from Russia. This platform is small enough to speak more or less openly, so here is a quick recap of what the last week looked like from the inside.
Putin declared mobilization: men with previous military experience must go to war, anywhere between 300k and 1 mil of them all over the country (the precise number is kept secret). On the first day, no clarification about categories was given, and hell broke loose. Amputees, elderly, people who never served in the army, parents of infants were drafted. Police would break into houses and catch people in subway.
The mobilization was not evenly distributed. In small towns and villages more people were drafted, to the point where protest started, especially in southern regions. In big cities it is less noticeable and smaller percentage of men are drafted, while more have the resources to flee.
The chaos subsided over the next few days. Some people are now excluded from mobilization, like students and workers of "crucial infrastructure", including financial sector. Most employers fight for their workers and try to keep them safe. Universities and scientific institutions do the same, draw all the resources they can.
From the first day, people started running. There is a humanitarian catastrophe on the border with Georgia, people have to pay to locals even to get in lines. Airplane tickets are sold out, ones that left are priced 10k$+. Estimates say that at least 200k left for Georgia and Kazakhstan that have land border with Russia. Better estimates do not exist at the moment, but it is compared to the Exodus by some.
Many people agree to fight and even buy ammunition for their own money to have better chances of surviving. Some of them are patriots, most of them seem to be obedient to whatever ones in power say.
Not really, I'll have to get into MSc program. Maybe more physics-centered, yes, but it's still science. Programming is more of a temporary job at this point.
Well this took a good while, eh. I finished my thesis! And started working as a full-time software tester the next day. Because aint no rest for the wicked, that and the experience and skills I can get.
It's an interesting experience after four years on university: to have a "real" job. And (for now) way less stressful than the studying used to be. There is a fixed workday, lunch, two (!) days off on weekends and plenty of time to do the job you were hired to do. And even a paycheck, not that large for now, but still eight times what my lab used to pay.
I hope this experience serves me well, even though now it breeds more trouble than it solves.
Lol. 1.5l is a lot, at least in terms of bathroom break time.
It's either tiramisu (although it's tricky to find a good one sometimes) or maple syrup bun with pecan nuts on top. And coffee is usually black, espresso, filter or lungo.
A cup of hot black coffee and some pastry make life a lot more tolerable. Add a cozy cafe with some good beats and you can almost feel alive.
Anyway, how have you been this week, my occasional audience? I promise a more elaborate story in a few days once I get enough sleep to recover and enough alcohol to forget the tomorrow's "intro to philosophy" test. This much may take a while to consume but then again, I am patient.
Ugh. So... Yeah. Fuck me, that wasnt an easy week.
...what do you mean its only Wednesday?
I have to clarify: it is not more difficult because of the information density (there was more material in some years), but because a) I have to write a thesis and b) teachers mostly have shitty time management and final tests for all subjects overlap on the last two weeks. I already had a day with two credits (one-on-one talk with a professor) and a test.
