/sci/ - Science & Math

>>11271462
>It makes the entire subject field moronic
Yes, it's pretty embarrassing for the professor who reviewed it, but actually the journal got tons of letters from PIs pointing out the obvious, not only that the method is famous and literally taught to children around the world, but also that issues with her version of it (hidden assumptions she failed to explicitly state, and no attempt to establish error bounds).
Believe it or not, scientists are not all one person.

>That means multiple people thought she'd unironically developed a new technique that hadn't been discovered before
But that's not the purpose of journals. Journals are a method of communication with the scientific community of a field, nothing more, nothing less.
It's not about whether the technique has been discovered before, it's about whether the journal's audience knows about it.
If you find a way to apply some technique from one field to another field in some way that's useful, that second field's scientists will find it useful and share it with their community.

Take a look at this article particularly the last section
https://www.quantamagazine.org/neutrinos-lead-to-unexpected-discovery-in-basic-math-20191113/

https://arxiv.org/abs/1908.03795v2
>In this survey we describe the many times that this identity, or variants thereof, have been discovered and rediscovered in the literature (with the earliest precursor we know of appearing in 1934).

>>10598861
Why aren't you going for a PhD?

>>10569507
yes, in the sense that the only "true" speed increase comes with improving clock speed.
minimizing memory transfer bottlenecks and parallelizing workloads yields substantial, but still marginal increases.
it could be argued that we've been stagnant for the last 30 years in the sense that we've been working primarily on a single technology (MOS).
the way forward is to find a new underlying technology that allows for even faster switching speeds.

>>10546327
Interesting perspective.
I've heard probably hundreds of students whine about their professor not giving them enough guidance, and I always remind them they're working on stuff that's hardly ever (or never) been done.
I've also heard professors complain about students expecting to be told everything.
I've certainly never heard a student complain that their professor is too hands-on except in the sense of "that guy actually makes us log our hours / sign into the lab at 8 am every day"
This must be field-related. What field do you work in?

>>10505787
this is a borderline /g/ question tbqh but anyways...

>crippling the time complexity
that's not how "time complexity" works

>messing up my hardware
your program will just run slow as balls. it shouldn't break anything (assuming your system has a proper cooling setup, etc).

anyways it is possible to efficiently crunch more data than the GPU has memory to store. you can hide the data transfer latency by performing it in parallel with computations.
They added an easy way to do this called CUDA Streams a few years ago iirc.
https://developer.download.nvidia.com/CUDA/training/StreamsAndConcurrencyWebinar.pdf
https://devblogs.nvidia.com/how-overlap-data-transfers-cuda-cc/

>>10478991
>expecting /sci/ to even know what a PLL is

>>10479111
i've actually gotten some good responses from it from time to time, it might take a few tries though