|
ultrafilter posted:Lot of selection effects and confirmation bias going on right now. sorry dr. filter
|
# ? Dec 6, 2021 21:37 |
|
|
# ? May 23, 2024 12:17 |
|
defmacro posted:part of the problem is there's a shitload of CS PhDs relative to other fields and they can't all get faculty jobs so they gotta end up somewhere. doesn't help that the code they wrote really only needs to work once for the paper. Hell, that's any academic programming. The standard of correctness for research code is "has to generate at least one result set, once, that looks superficially plausible to a credulous supervisor and two or three reviewers of varying skepticism." That's it. It's a really bad state of affairs.
|
# ? Dec 6, 2021 22:43 |
|
Absurd Alhazred posted:That might explain it - I'm not a CS PhD. I'm a socially inept know-it-all for completely different reasons
|
# ? Dec 6, 2021 23:14 |
|
raminasi posted:Hell, that's any academic programming. The standard of correctness for research code is "has to generate at least one result set, once, that looks superficially plausible to a credulous supervisor and two or three reviewers of varying skepticism." That's it. It's a really bad state of affairs. it's definitely better in some fields. it seems like statisticians slap their R monsters on CRAN pretty readily. and security is starting to improve at the big paper venues by giving more credit for open data/code. but yeah the bar is super low.
|
# ? Dec 6, 2021 23:41 |
|
There's a movement to improve code quality and the reproducibility of analyses. Right now it's mostly concentrated in psychology, but it's having an influence on other fields. Progress is slow because science literally advances funeral by funeral, but there's momentum building up and we will eventually get to a better state. (Computer science is one of the worst fields, though, because they've never really come to terms with being an actual science. Someday.)
|
# ? Dec 6, 2021 23:44 |
|
ultrafilter posted:(Computer science is one of the worst fields, though, because they've never really come to terms with being an actual science. Someday.) Computer science is one of the best fields because it has Edsger W. Dijkstra, who wrote all his papers in longhand and cited nothing, and that’s a hero move.
|
# ? Dec 7, 2021 00:39 |
|
lifg posted:Computer science is one of the best fields because it has Edsger W. Dijkstra, who wrote all his papers in longhand and cited nothing, and that’s a hero move. That's computing science. Totally different. More seriously, though, CS is weird since the mathy parts have, well, mathematical rigor, but for anything that's more engineery it's basically impossible to really do controlled evaluation of ideas.
|
# ? Dec 7, 2021 00:43 |
|
CS is a very math-y major that desperately wants to be seen as an engineering discipline like its bigger brothers EE and ME.
|
# ? Dec 7, 2021 00:46 |
|
There are some people doing rigorous experimental work in hardware and HCI. You could theoretically do that in ML as well, but it's pretty rare. I think the only subfield where it's really tough is software engineering.
|
# ? Dec 7, 2021 00:48 |
|
ultrafilter posted:There are some people doing rigorous experimental work in hardware and HCI. You could theoretically do that in ML as well, but it's pretty rare. I think the only subfield where it's really tough is software engineering. I was thinking mostly systems. Like I would read a paper making a reasonable case about how their system using idea X is an improvement over a previous system because of X, and wonder if it's really X doing it or just a bunch of hundred of others small decisions they have made when implementing this really complicated thing that made theirs better. (And of course proper answer in systems are highly dependent both on workload and various performance ratios of things). Edit: I imagine software engineering could learn something from social sciences; it s, after all, about humans F_Shit_Fitzgerald posted:CS is a very math-y major that desperately wants to be seen as an engineering discipline like its bigger brothers EE and ME. Depends a lot on individual group. Like I don't think anyone using denotational semantics would pick engineering over math.... OddObserver fucked around with this message at 01:01 on Dec 7, 2021 |
# ? Dec 7, 2021 00:53 |
|
OddObserver posted:Depends a lot on individual group. Like I don't think anyone using denotational semantics would pick engineering over math.... Yeah, I'll admit that I'm looking at it from the perspective of someone who studied CS in a mostly software-oriented context. I don't know all that much about them but I'd imagine that ML and robotics are 'harder' engineering than coding.
|
# ? Dec 7, 2021 00:56 |
|
F_Shit_Fitzgerald posted:ML...'harder' engineering than coding.
|
# ? Dec 7, 2021 01:00 |
|
F_Shit_Fitzgerald posted:Yeah, I'll admit that I'm looking at it from the perspective of someone who studied CS in a mostly software-oriented context. I don't know all that much about them but I'd imagine that ML and robotics are 'harder' engineering than coding. Robotics sure, ML haha... no.
|
# ? Dec 7, 2021 01:02 |
|
also this https://github.com/elfring 404's now - is this the end of an era?
|
# ? Dec 7, 2021 01:15 |
|
all those pull requests lost in time... like tears in rain
|
# ? Dec 7, 2021 01:20 |
|
repiv posted:all those pull requests lost in time... like tears in rain Pull Request.... denied.
|
# ? Dec 7, 2021 02:29 |
|
https://twitter.com/TeaStats/status/1467785202567200768 Just in case R's not inconsistent enough for you.
|
# ? Dec 7, 2021 05:33 |
|
F_Shit_Fitzgerald posted:I'd imagine that ML and robotics are 'harder' engineering than coding. I spent four hours QCing a dataset yesterday afternoon.
|
# ? Dec 7, 2021 15:17 |
|
Supporting both 0 and 1 based indexing is a good way to gently caress up your language (or API, looking at you MKL)
|
# ? Dec 7, 2021 21:42 |
|
Beef posted:Supporting both 0 and 1 based indexing is a good way to gently caress up your language (or API, looking at you MKL) IIRC the original BASIC had an OPTION BASE 0/1 command, checks out.
|
# ? Dec 7, 2021 22:05 |
|
Zopotantor posted:IIRC the original BASIC had an OPTION BASE 0/1 command, checks out. This was true through VB6
|
# ? Dec 7, 2021 22:11 |
|
Just last week my team encountered a defect around VB.Net's array initialization syntax. (You need a Dim foo(-1) as Integer for an empty array if you were wondering)
rarbatrol fucked around with this message at 00:18 on Dec 8, 2021 |
# ? Dec 8, 2021 00:16 |
|
Beef posted:Supporting both 0 and 1 based indexing is a good way to gently caress up your language (or API, looking at you MKL) Old versions of Perl had a configurable variable which could make it use 2-based indexing or whatever you wanted.
|
# ? Dec 8, 2021 02:20 |
|
R's one-based indexing has never been a problem for me. Occasionally if you have to manually walk through a vector by index you'll have to remember to write "for (i in 1: length(v)" instead of "for (i in 0:(length(v)-1)" and that's the rare mode of iteration anyways. It's also occasionally useful to take the zeroth index of a vector to get an empty vector of the same type.
|
# ? Dec 8, 2021 03:26 |
|
I played around with the index0 package a bit. If you take a non-existent index from a standard vector you get an empty vector, but if you do the same with a 0-indexed vector, you get NA.
|
# ? Dec 8, 2021 04:08 |
|
rarbatrol posted:Just last week my team encountered a defect around VB.Net's array initialization syntax. (You need a Dim foo(-1) as Integer for an empty array if you were wondering) Hah, looking this up led me to learn of the ReDim statement. VB is full of weird little things like that.
|
# ? Dec 8, 2021 05:02 |
In Pascal (or at least the Delphi variant of it), you can define an integer type with some arbitrary range, and declare an array indexed by that range type. You can also declare an enumeration type with named values and declare an array of that enumeration. I think the syntax goes something like this: code:
|
|
# ? Dec 8, 2021 09:03 |
|
raminasi posted:VB is full of weird little things like that. Or maybe it's me who considers them weird because I expect arrays to be packed like C-arrays. Wipfmetz fucked around with this message at 14:43 on Dec 8, 2021 |
# ? Dec 8, 2021 14:32 |
|
raminasi posted:Hah, looking this up led me to learn of the ReDim statement. VB is full of weird little things like that. I thought at one point or another I'd used every obscure piece of VB.NET syntax, from Call to Static to Shadows, but it turns out I'd never even seen Erase. The old VB garbage collector must have sucked bad if you couldn't just do 'someArray = Nothing' and trust that it would GC the elements as well, instead having to invoke a special statement.
|
# ? Dec 8, 2021 22:21 |
|
nielsm posted:In Pascal (or at least the Delphi variant of it), you can define an integer type with some arbitrary range, and declare an array indexed by that range type. You can also declare an enumeration type with named values and declare an array of that enumeration. Yup, that's standard Pascal.
|
# ? Dec 8, 2021 22:25 |
|
I miss Pascal.
|
# ? Dec 8, 2021 22:55 |
|
Absurd Alhazred posted:I miss Pascal. I've never actually written a Pascal program, but I came up programming in C on the old Mac, and reading documentation for the Macintosh Toolbox which was all assuming you were using Pascal or assembly. I got real used to length-prefixed strings.
|
# ? Dec 8, 2021 23:18 |
|
NihilCredo posted:The old VB garbage collector must have sucked bad if you couldn't just do 'someArray = Nothing' and trust that it would GC the elements as well, instead having to invoke a special statement.
|
# ? Dec 9, 2021 03:46 |
|
Absurd Alhazred posted:I miss Pascal. I never programmed much in it, but having had to TA intro CS classes in Java I feel like it's missed badly as a teaching language (maybe adjusted to have better strings).
|
# ? Dec 9, 2021 04:17 |
|
nielsm posted:In Pascal (or at least the Delphi variant of it), you can define an integer type with some arbitrary range, and declare an array indexed by that range type. You can also declare an enumeration type with named values and declare an array of that enumeration. That is the sanest way I've seen to deal with the indexing base problem. APL has the quad-io construct which changed the behavior of all subsequent operations, introducing a global effect and wrecking modularity. (not an APLer, just listen to ArrayCast)
|
# ? Dec 9, 2021 09:27 |
|
ccode:
|
# ? Dec 9, 2021 10:20 |
|
seems fine? We called then yes, no, cantsay though
|
# ? Dec 9, 2021 12:05 |
|
Tei posted:c The canonical third boolean is FileNotFound.
|
# ? Dec 9, 2021 12:47 |
|
True = 1, False = 2, Nice = 69, Dope = 420
|
# ? Dec 9, 2021 15:33 |
|
|
# ? May 23, 2024 12:17 |
everything in life can be simplified down to two options, true, false, and null.
|
|
# ? Dec 9, 2021 16:12 |