jnxx

59 ms·

541.
by jnxx 6y ago
> Although I've never been to Berlin, I can confirm that in USA they pay no attention to cyclists. The best advice is to never be beside one. Only attempt to pass on the right if they are completely stopped with other completely st
542.
by jnxx 6y ago
> After all, I'm wasting 2 months of my PhD for the marriage of my own results with known results which -in principle- could have been done within one day if the code base would allow for it. Sounds like it is quite good science to
543.
by jnxx 6y ago
In many parts of scientific research, researchers are, to stay in your metaphor, more travelers using a map, than map makers. Of course, it is a difference whether you make a clinical study on drugs, and use a pocket calculator to compute
544.
by jnxx 6y ago
> Race conditions aren't undefined behavior in C/C++. Data races are. You are right with the distinction, I had data race conditions in mind. Race conditions can well happen in a correct C/C++ multi-threaded program in the
545.
by jnxx 6y ago
> Monte-Carlo can and should be deterministic and repeatable. That's a nitpick, but if the computation is executed in parallel threads (e.g. on multicore, or on a multicomputer), and individual terms are, for example, summed in a ra
546.
by jnxx 6y ago
There is something I need to add, it is a subtle but important point: Non-determinism can be caused by a) random seeds derived from hardware, such as seek times in a HDD controller, which is fed into pseudo random number (PRNG) generation.
547.
by jnxx 6y ago
With the caveat that while in some cases, like computational science, numerical analysis, machine learning algorithms, computer-assisted proofs, and so on, details of the code could be crucial, in other cases, they should not matter that mu
548.
by jnxx 6y ago
If nothing else, it helps debugging code which tries to reproduce your findings.
549.
by jnxx 6y ago
> I'm an accelerator physicist and I wouldn't want my code to end up on acceleratorskeptics.com with people that don't understand the material making low effort critiques of minor technical points. I'm here to turn ou
550.
by jnxx 6y ago
Well, where do you download the hardware ? ;-)
551.
by jnxx 6y ago
Python is an extremely bad example. Try twenty years old Common Lisp code. Or Fortran.
552.
by jnxx 6y ago
> Spack/Guix/nix are the best solution in theory, but they come with a host of other problems that made them less desirable. I would be quite interested to learn more what these problems are, in your experience. I've only
553.
by jnxx 6y ago
I am referring to this blog post: https://lockdownsceptics.org/code-review-of-fergusons-model/ It says, word-by-word: > Clearly, the documentation wants us to think that, given a starting seed, the model will alway
554.
by jnxx 6y ago
Well done!
555.
by jnxx 6y ago
Konrad Hinsen is an expert in molecular bioinformatics and also has significantly contributed to Numerical Python, for example, and has extensively published around the topic of reproducible science and algorithms - see his blog. The fact t
556.
by jnxx 6y ago
> CommonLisp Common Lisp has fantastic long-term stability. I think that deserves more recognition, as Common Lisp is often almost as fast as C, but is (by default) not riddled with undefined behavior. It would be superb if Rust could ta
557.
by jnxx 6y ago
> It would be considered a dick move to criticism someones code, if the code was proven to produce the correct result. Formal proof is much much harder than making code understandable and reviewable. It can be done but it is not easy, an
558.
by jnxx 6y ago
I don't think that would be any problem (why should it?). Code exhibiting undefined behavior is a different kettle of fish...
559.
by jnxx 6y ago
And it seems that the people from Imperial College have done that with their epidemiological simulation. What critics claim is that their code produces non-deterministic results when given deterministic input and random seeds, i.e. that the
560.
by jnxx 6y ago
> If you cannot the release the code because the code is too ugly or untested or has bugs, how do you expect anyone with the right expertise to assess your findings? Yes, that should be this way. Also all cases where some company resear
561.
by jnxx 6y ago
edit: please read the grandchild comment before going off on the idea that some random programmer on the Internet dares to criticize scientific code he does not understand. What is crucial in the argument here is indeed the distinction be
562.
by jnxx 6y ago
This. In the last years, conventional software engineering has in many cases experienced an explosion in complexity which will make very very difficult to maintain stuff in the long run. This only works because over 90% of startups go bust
563.
by jnxx 6y ago
Companies with code in production have a short-term real incentive to keep that code running. This is different from code from research projects, which is on many cases just run a few times, and in other times, written by somebody who has,
564.
by jnxx 6y ago
Open textual formats for data and open source application and system software (more precisely, FLOSS), are just as important. Imagine that x86 - and with it, the PC platform - gets replaced by ARM within a decade. For binary software, this
565.
by jnxx 6y ago
> package your code into a docker docker is not a general solution for this. What is needed is a way to re-generate everything from source and from scratch.
566.
by jnxx 6y ago
As long as it does not use some CUDA hardware which is using tensorflow and Numba which is using a version of llvmlite which does not support Python2 any more..... This isn't a theoretical example.
567.
by jnxx 6y ago
Even more fundamental is that there is no maintenance budget for important scientific libraries and tools. Somebody wrote them as part of their job, and the person who wrote it, is now working somewhere else.
568.
by jnxx 6y ago
Guix is arguably better.
569.
by jnxx 6y ago
Very related to this, see also Hinsens blog post: http://blog.khinsen.net/posts/2017/11/16/a-plea-for-stabilit... I think that GNU Guix is extremely well-suited to improve this situation. Also, one could
570.
by jnxx 6y ago
And that's just equivalent to a human lifting a chair without hurting himself.

More ›