Thanks Day, very cool!

[NynJa]

you two are acting like a couple real scorpios rn

Dont mix me up with those two nerds.

Unless theres a difference between tober-Scorps and novembro-scorps

[BerenTebogo]

That's gotta suck ass. I was speaking from a position of never having to do almost anything by hand. I firmly believe that (in my situation) knowing how to do it with software and why you do it is more important than being able to do it by hand. It's possible I'm just bitter about the shit I don't actually get to use.

but how do you know if you wrote the software right and the results can be trusted?

like what if you have an off-by-one error somewhere and it doesn't really pop up all the time?

[BerenTebogo]

or a sign error that only happens when you have a leading 0 or something

[BerenTebogo]

I can code up, down, and sideways and have been writing supercomputer simulations for a number of years and the amount of times I fucking make mistakes that I only find after I think I'm "done" and doing spot testing is fucking huge

[Day]

Dont mix me up with those two nerds.

Unless theres a difference between tober-Scorps and novembro-scorps

im a scorpio too, im just fuckin w sath

[NynJa]

im a scorpio too, im just fuckin w sath

Hell yeah

[NynJa]

Adrock is also a scorpio

Wait you listen to country music nvm

[Demosthenes]

but how do you know if you wrote the software right and the results can be trusted?

like what if you have an off-by-one error somewhere and it doesn't really pop up all the time?

It's a massive fear and I do a lot of checking along the way, iPython notebooks are good for data research like that. I work in segments and when I get to the end I codify a pipeline and push other stuff though.

Don't get me wrong it's important that there are no mistakes in your work, but statistics is about generalizing and not solving 1:1. 1:1 models are super bad and if I get a model that gives me above 85% representation, I immediately distrust it.

[BerenTebogo]

It's a massive fear and I do a lot of checking along the way, iPython notebooks are good for data research like that. I work in segments and when I get to the end I codify a pipeline and push other stuff though.

I don't get me wrong it's important that there are no mistakes in your work, but statistics is about generalizing and not solving 1:1. 1:1 models are super bad and if I get a model that gives me above 85% representation, I immediately distrust it.

idk I guess this is the difference between specializations

I can't imagine presenting any results or code to anyone without as much time testing as it took me to come up with and write it.

[BerenTebogo]

then again most of the code I write these days is used to keep a space telescope that's 20 years old operational, productive, and generating good science results

mistakes aren't really allowed

[Demosthenes]

idk I guess this is the difference between specializations

I can't imagine presenting any results or code to anyone without as much time testing as it took me to come up with and write it.

SAS and most Python data science libraries are hardened enough to prevent you from fucking yourself in the ass. I don't know what MatLab (or octave) is like these days, but I'd REALLY be anal about anything I wrote in it and would probably spend ages doing full audits after every working day. I'd probably do a lot more by hand if I used MatLab.

[Tokitoki]

Adrock is also a scorpio

[Day]

life changing album for me

[Demosthenes]

then again most of the code I write these days is used to keep a space telescope that's 20 years old operational, productive, and generating good science results

mistakes aren't really allowed

Yeeaaah, dropping the machine learning part, my job is basically creating formulaic representation of data and identifying hidden trends. Yours demands complete accuracy, mine demands an acceptable generalization. Acceptable can vary a lot. A solid 95% is the work of a deity in my specific area of work, but in statistics for medicines and trials for example a 95% can be absolutely disastrous.

[BaneTheBrawler]

medicines and trials, you say. interesting. i might actually be generating some of that data. neat.


edit: aww, misread. thought you meant you worked on that specifically. still, neat the way your field connects to others.

[Demosthenes]

medicines and trials, you say. interesting. i might actually be generating some of that data. neat.


edit: aww, misread. thought you meant you worked on that specifically. still, neat the way your field connects to others.

Stats is broad as fuck and I fucking enjoy it, but holy fuck, I'd hate to be in the pharmaceutical industry doing it. I'd probably neck myself over the guilt alone.

[Demosthenes]

I'd probably neck myself

Do it anyway you fucking coward!

[Anoat]

RIP Xbox, MS rolled out an update then servers went down so they have essentially bricked every console that was connected to the Internet at the time. I’m sure a fix is coming but people hella mad right now.

[dasva]

Nerds

[Zealot]

india is getting right with the world at the same time that they're building a large interferometer to measure gravitational waves so tbh my vote is to make LIGO-Afghanistan and see if things work themselves out

Gravitational waves are the biggest joke of legitimate science communities.

Sent from my SM-N960U using Tapatalk