« first day (2705 days earlier)      last day (2128 days later) » 

12:57 AM
@doork hello there!
 
messing around with neural networks :D
 
nice
 
that's ~4k parameters needed to represent the neural network
 
I've also been messing around with neural networks recently.
Different kind, though.
 
what kind?
 
1:04 AM
Biologically-accurate simulations... simulating reinforcement learning by modeling dopamine bursts in the striatum and measuring response times in the thalamus.
 
I really dislike the name neural network
to be honest
what you're doing is neural networks
what I'm doing is differentiable programming :P
I don't really care about 'neurons' or whatever
I just want a sufficiently complicated universal approximator I can differentiate w.r.t its parameters
 
1:24 AM
@PhiNotPi o/
 
How's life?
 
pretty good
how've you been?
 
And yeah, I think there's a pretty big gap between the two categories of neural network. I don't really know a better name for them though.
I've been good... still doing undergrad research this summer (which is what I've been talking about).
 
sounds fun
how long have you been working on it?
 
:O It's a doorknob!
o/
 
1:33 AM
@PhiNotPi differentiable programming
 
This is the same project as last summer, I've been at it these past couple weeks.
 
we're programming functions that we can differentiate w.r.t parameters
that's the cruz of 'neural networks'
 
@DJMcMayhem hey! o/
 
that they're differentiable and that we can follow the gradient to opimize them
@Doorknob hey o/
 
2:17 AM
Looks like I have not visited the featured tab for a while...
 
0
A: Sandbox for Proposed Challenges

GammaGamesChallenge n=2πr Given an input, n, print a circle with the circumference of n characters. Rules The characters for the circumference of the circle can be any non-whitespace character Shapes don't have to be an amazing circle, kinda roundish shapes are still valid The circle can have a small ...

 
 
2 hours later…
4:18 AM
CMC: Given content of stdin in the form of print the smaller number between <integer> and <integer>, do it.
Example I/O: print the smaller number between -3 and 2 -> -3
 
4:35 AM
0
A: Sandbox for Proposed Challenges

user71546Count "n Queens Problem" solutions in shortest code code-golf I know there are a handful of questions about the "n Queens Problem" or anything similar, but I'm quite surprised that none of those are code-golf challenges. As a result, I would like to post this challenge as a code-golf challeng...

 
5:10 AM
@user202729 Japt, 18 bytes
 
6:01 AM
@user202729 Jelly, 9 bytes
 
@user202729 why €
 
I'm not supposed to remember vectorization rules of every atoms right
Actually the CMC is intentionally designed so that Hexagony get an advantage (? in Hexagony auto ignore non digits)
But it still loses to Jelly ...
 
Wait what is
Uh sort...
 
doesn't work
if 2 negatives
 
6:09 AM
doesn't work for 3 and 12
 
my bad
@user202729 J, 16 bytes: echo<./_".1!:1]3
 
I never used Numbers (dyadic "....)
Nice.
And Jelly doesn't have "exception catch" (adverse)
 
adverse is so cute
 
@user202729 QuadS, 14 bytes ⌊/⍵[-\d]+⍎⍵M
 
6:28 AM
Looks like that most solutions take advantage of the fact that all other words don't have any digits (except the Jelly solution)
@Dennis What's wrong with this?
 
@user202729 Sounds like a sensible approach.
 
There are some (a lot?) invalid/comment/not-serious answers, but the question looks fine. (although a bit poorly defined)
 
6:57 AM
CMC: For fixed integer C>1, find integer n>0 such that 14×n+3×n×ⁿ√C is minimum.
(background: Preproc and nested define with a target)
 
yesterday, by mınxomaτ
GitHub is fine. Microsoft's VSTS (the real Gitlab competitor for "real" enterprises) is also pretty great. VSTS build system e.g. is awesome, comparable to e.g. Buildkite.
Well, today Microsoft published an update to VSTS which changes the UI layout to be the same as GitLab. It's actually much nicer than the weird UI before it.
 
 
1 hour later…
8:28 AM
@user202729 the userscript creates the leaderboard for all challenges github.com/vihanb/PPCG-Design
 
 
1 hour later…
9:29 AM
After finding meta I can find this, but it is 2 years older than the lock.
 
 
2 hours later…
11:10 AM
0
Q: A nicely-spaced ascii spiral

AJFaradayConsider this spiral ########### # # ####### # # # # # ### # # # # # # # # * # # # # # # # ##### # # # ######### Starting in the centre: The first line (upwards) has 3 characters. The second line has the same number of characters(3) Next, we add two chars (5) for the next two si...

 
11:28 AM
@user202729 English, 0 bytes
 
12:39 PM
@Neil Two questions regarding Charcoal: 1. Why isn't there a MoveBack command? I know there is a Move and Jump, but let's say you are printing a spiral pivoting every iteration, a MoveBack in the opposite direction would be useful. 2. If you've used InputNumber() and want to use the same input again, why does q not remember it's a number? I now have to use Cast(q) if I want to re-use the input-number..
 
1:34 PM
@KevinCruijssen not Neil, but 2. InputNumber(q)
(note: you can't use that as an expression)
 
1:57 PM
@EriktheOutgolfer Hmm, I'm not sure I completely understand. Or at least, it's not working as I expected when I read your comment. How would you use this InputNumber(q) in this code of mine?
 
like this
-1 byte
 
@EriktheOutgolfer Ah ok, thanks. You first save the InputNumber in q before actually using q multiple times. I has something like this: For(InputNumber(q)), but that didn't work.
 
because InputNumber(q) isn't an expression
it's a statement
 
I see. TIL, thanks. :)
@EriktheOutgolfer Seems it's sometimes still shorter to not save it even though using it multiple times, though. 17 bytes with one explicit and one implicit cast for other two usages, and 18 bytes by saving it as Number in variable q.
 
1
A: Sandbox for Proposed Challenges

LaikoniConvincingly Fake Compression of Random Data cops-and-robbers As you may know, it is impossible to write an compression method that takes strings of length \$n\$ and returns strings of length \$n - 1\$. This can be proven by a simple argument: Each compression function must be a bijection, becau...

 
2:13 PM
@KevinCruijssen that's because you use it only twice
 
@KevinCruijssen (not the author of Charcoal, but) 1. unfortunately the only pivot-relative movement command available is Print("\n"); although I've used that to great effect myself previously. 2. (continued...)
(...continued) The predefined variables were originally really predefined, so that they were set before your program started. Nowadays I think they prompt if there weren't enough values passed to the program, but they're still strings. However, you can cheat and pass your input in JSON format, in which case the variables use the type from JSON.
 
@user202729 It got migrated from SO, then reworded completely. Most of the 60+ answers are invalid.
 
2:44 PM
ugh, I need a grace period for comments so that I can fix my "now arms" to "arms now" without anyone noticing
 
@Neil I can edit that for you
If you give me a link
 
Because the now arms overlap you need to make them an extra # longer. — Neil 2 mins ago
 
@Neil both comments and messages have a time limit on editing altogether
 
Because the arms now overlap you need to make them an extra # longer. — Neil 2 mins ago
 
but now everyone noticed
 
2:46 PM
well, they noticed anyway because of the pencil
 
whistles innocently
3
 
omg, the question precisely
thanks
 
@FrownyFrog Use unions instead. Try it online!
(Solves 1, not 2.)
 
3:34 PM
@FrownyFrog *(a[2]-1) silences the warning
 
Actually in most cases it's guaranteed to be correct.
Just don't let the compiler know that you invoke UB and you will be fine.
 
@Dennis *(a[2]-1) fixes that too ;-)
((int*)a)[i] = i; also seems to work
 
That's a band-aid solution though and might mysteriously break all of the sudden.
gcc warns about a[0][i] = i; when using *(a[2]-1) with optimizations.
@Neil That's essentially what the union does, only in a safe way.
 
I always thought that reading from a union member that wasn't the last one written was still UB
 
3:48 PM
Huh, you're right.
 
CMC: given an image with a white background containing coloured rectangles, like this, remove all the red rectangles
for the example given above, the following is the correct output i.stack.imgur.com/eIsA4.png
 
MS paint, 5 clicks :P
 
want to make an esolang out of that :)
I really want there to be more golflangs
 
Red as in #FF0000? How close do we need to match colors?
 
4:03 PM
yes #ff0000, the sample will only contain colours from #{0,f}{0,f}{0,f}
 
PowerShell will be somewhere between 300 to 400 bytes, looping through each pixel and if it's red setting it to white, then re-saving the image.
 
oh right, you can simply replace each red pixel with a white one
 
Sure
 
how do you think the challenge can be improved?
(I'm planning to sandbox it)
 
@Cowsquack Do you want to prevent solutions like that?
 
4:05 PM
@Dennis But in that case the compiler know that UB is invoked.
Replacing all a[b] with *(a+b) appears to make GCC quiet.
 
What if each rectangle was white with a colored border, and your goal was to change the color inside of the red rectangle?
 
I hoped to, but now I don't know how to do that (or I can simply create a new challenge all together to replace red pixels with white ones)
@DJMcMayhem that is a nice idea
 
@Cowsquack It's just a golflang with builtin input parsing/generating image format.
 
@user202729 I don't think failure to print a warning makes the code much safer.
 
@Dennis GCC does not recognize it's UB means that it won't aggressively optimize.
 
4:07 PM
@user202729 and image manipulation, animation, and other stuff
 
And I'm 99% certain that the "natural" way to do it coincides with the expected result.
@Cowsquack Animation is just GIF output format. Image manipulation is just ASCII art, replace 'char' with 'color'.
I expect that slightly modified Charcoal (or other ascii golflang) would perform reasonably well.
 
@Neil man gcc talks about type-punning. The union thing seems to be allowed by gcc, even if it's UB according to the standard.
 
@DJMcMayhem maybe also make the rectangles have variable dimensions (and actually rectangles rather than simple squares >_>) and maybe rotations
my original idea was to use circles, so those could be used instead
 
If you're looking at a pixel by pixel level, how do you define what a diagonal line should look like?
 
@user202729 Recognizing and warning about are different. With default switches, gcc 4 doesn't warn about this, but it still produces unexpected output.
 
4:15 PM
good question DJ
 
@Dennis That explains for the non serious contender/pseudocode answers.
Initially it asks for shortest algorithm, then changed to code golf.
 
I suppose I can leave the choice of aliased/anti-aliased input to the answerer
 
@Dennis AFAIK if GCC recognizes an (exploitable) UB it will give warning with -Wall.
 
I can't imagine why anyone would prefer aliased
 
@user202729 Even if that's true, why write code that might break in the next version of gcc?
 
4:19 PM
@Dennis If you don't need it to work until the next version.
Anyway, why would somebody write that code in the first place.
 
4:35 PM
@Cowsquack I'm not sure that filling a rectangle (or circle) would be that much more difficult. Have two Boolean values, one for x and one for y positions, then when you're scrolling pixel by pixel flip the values to $true if you're "in" the bounding box. If both are $true, flip the pixel color, else leave it the same.
 
5:00 PM
@Cowsquack If input and output can be in PPM format (which is indeed meta consensus), it amounts to a simple string search-and-replace.
 
@AdmBorkBork fair point, then I'll have to think about that to make the challenge more challenging
@Adám in that case I can disallow ppm
*explicitly disallow ppm, and restrict image formats to just a few
 
@Cowsquack Want to disallow bmp too?
 
or maybe I should enforce output to be displayed "graphically"
I don't know how to do the same for input
 
@Cowsquack Maybe, but that makes online testing very hard.
 
hmm, or I should simply not disallow them, after all, I'm sure the languages with builtin png libraries treat them similarly to those that treat other formats
(and would hence give an advantage to the ones with png and such libraries)
@Adám I could wait until tio gets graphical output :D
 
5:11 PM
@Cowsquack Yeah, I think leaving default I/O as-is is the best. It also lets people explore "clever" solutions.
@Cowsquack Sure, when is that?
 
@Dennis ^
 
Currently not very high on the list. I won't implement graphical output with client-side permalinks, at least not without talking to a lawyer.
There would be no way to remove copyrighted or obscene images.
 
5:27 PM
@Adám What is this for?
 
2 hours ago, by Cows quack
CMC: given an image with a white background containing coloured rectangles, like this, remove all the red rectangles
@Dennis How is that any different than letting people open copyrighted or obscene images in your image viewer application? Sure if you had server-side permalinks, you might get in trouble, but client-side? The user is supplying all the data and nothing remains on your server!
4
 
5:44 PM
Dennis haven’t spoken to a lawyer yet
 
0
A: Sandbox for Proposed Challenges

Luis felipe De jesus MunozReplace with sequence Given a pattern (string or array format) of 1 and 0 : [0,1,1,1,0,1,1,0,0,0,1,1,1,1,1,1] The tasks is to replace consecutive 1 with an ascendant number sequence starting at 1. Example: given [0,1,1,1,0,1,1,0,0,0,1,1,1,1,1,1] the output should be [0,1,2,3,0,1,2,0,0,0,1,2,3...

 
I'm assuming he doesn't want to take any risks.
 
6:08 PM
0
Q: Smallest region of the plane that contains all free n-ominoes

Peter KageyOn Math Stack Exchange, I asked a question about the smallest region that can contain all free n-ominos. I'd like to add this sequence to the On-Line Encyclopedia of Integer Sequences once I have more terms. Example A nine-cell region is the smallest subset of the plane that can contain all t...

 
6:48 PM
I feel like I should respond to this, but I just don't have it in me.
 
The question will probably be closed anyways
 
I assume the comments will be purged if nothing else. The question itself seems factually answerable enough.
 
7:09 PM
@Geobits aw i'm sure your response would be a gem
he set you up something fierce
 
Yeah lol
After browsing the profile out of curiosity, it's taken me quite a bit of restraint to not serially downvote
 
7:27 PM
@Geobits Well. That was an entertaining couple of minutes.
 
I aim to please
 
7:48 PM
0
A: Sandbox for Proposed Challenges

Cat WizardVorogomlok complexity The kolmogorov complexity of a string is the length shortest program that outputs that string. A similar concept that I made up is Vorogomlok complexity. The Vorogomlok complexity of a sequence is the smallest number of terms in the sequence (consecutive and starting at 0...

 
8:23 PM
@NewSandboxedPosts @CatWizard I really like this one
 
Does it make sense? I feel like it might be confusing
 
Let me just read it again a couple times
 
It makes sense to me. I read through it once
 
I'm also realizing that the stuff about Vorogomlok complexity is probably not technically necessary to understand the challenge.
 
'Nother profile picture?
 
8:25 PM
Yes
 
@CatWizard Actually, I'm confused about something... Are cops defining a full sequence or just the first n terms?
 
Their program defines a full sequence, then they choose n.
 
OK, so then cracks must write shorter code that outputs any sequence where the first n terms are equal?
 
Yes
 
OK, that's clear now.
 
8:30 PM
Well it needs to be different than the original as well.
 
Sorry, yeah
 
Was revisiting this and had some thoughts:
4
A: Sandbox for Proposed Challenges

quartataRegex Golf Generators cops-and-robbers kolmogorov-complexity regex Challenge: Cops: The cops must post a 150 byte or less program in any language that outputs between 20 and 200 strings of printable ASCII (this excludes newlines), half of them "match" strings and half "don't match" strings. Y...

 
@CatWizard I think the example with id and ^2 confuses the task slightly because it looks like you want to find shorter code with a matching sequence rather than shorter code with a partially matching sequence
 
Leaving aside the fact that it's kinda poorly written
 
@DJMcMayhem Ok, yeah. I might want to take out the example.
 
8:32 PM
In the original spec, cops that weren't cracked didn't get any points at all (but self cracks were legal). The point was to force cops to make something that is crackable
That isn't the case now -- uncracked solutions get points as if someone had posted a crack that hard coded each string
Im not sure how I feel about this
Also, are the byte limit/number of string limits too arbitrary? I don't remember why precisely I picked them
 
@Adám Logically, it isn't. Laws aren't always logical though, especially here.
 
@DJMcMayhem I rewrote it quite a bit.
 
And laws aside, I'd like visitors to be confident they won't see something disturbing when they click a permalink.
 
@Dennis Fair enough. But are you not afraid people will use TIO to render copyrighted or obscene ASCII art?
0
Q: Can I used an ASCII generated version of a copyrighted image in my game?

CaimenI'm going to guess the likely answer is no, but I'm interested to know more. So let's say I have an image of James Bond from Goldeneye. Then let's say I run it through ASCII art generator such as this http://picascii.com/ and it produces the following ASCII code made for the web as such (can...

 
You can get a lot more graphic with images.
 
8:38 PM
@CatWizard That's an improvement
 
But from the legal standpoint, you're right; client-side permalinks may actually be safer for me/TIO.
 
@Dennis Yeah, I doubt the makes of your browser could get sued for allowing you to click on the link inside this TIO link:
 
@Adám What even is that?
 
@DJMcMayhem Just a "URL" that resolves to an image, but all the data is client side.
 
It's not quite the same. The client-side permalink is interpreted by the TIO frontend and sent to the TIO server, which replies with an actual image.
 
8:47 PM
No, what is the image?
 
@DJMcMayhem The icon on a new toolbar button in Dyalog APL version 17.0's Windows IDE. It is supposed convey "boxing", i.e. displaying [[[1],[2]]] as
┌─────┐
│┌─┬─┐│
││1│2││
│└─┴─┘│
└─────┘
instead of ` 1 2 `.
(How do I have leading and trailing spaces in chat markdown?)
@DJMcMayhem Do you think it conveys the idea well? If not, any ideas for improvement?
 
I think it conveys the idea pretty well
@Adám BTW, still waiting on a CMC answer :P
 
@DJMcMayhem Oh, I'm sorry. I did make two APL solutions but both were awfully long.
@Dennis The client-side "url" is interpreted by the browser and sent to the window manager as an actual image.
 
9:54 PM
@ASCII-only how do i make a sqaure in charcaol. This is my code but idk how to get the index of the loop
@ASCII-only i think i am refer to wrong loop variable? not sure how to do in verbose mode docs don't say
 
 
2 hours later…
11:40 PM
are there any other good units of time other than the oppressive second/minute/hour/day/year hierarchy
asking for my new country i'M starting...
OK this was an interesting read:
The traditional Chinese time systems refers to the time standards for divisions of the day used in China until the introduction of the Shixian calendar at the beginning of the Qing dynasty. == Han-era system == The third chapter of the Huainanzi outlines 15 hours during daylight. These are dawn (晨明), daybreak (朏明), morning (旦明), early meal (早食; 蚤食), feast meal (宴食), before noon (隅中), noon (正中), short shadow (少还; 小還), drum time (铺时; 餔時), long shadow (大还; 大還), high setting (高舂), lower setting(下舂), sunset (县东; 縣車), dusk (黄昏), rest time (定昏). These are correlated to each hour from 06:00 to 20:00...
 
AMA
 
They're still derived from Earth days, of course
But it, like, makes more sense because it's more base-10 oriented rather than base-60
@LeakyNun is this still used anywhere?
It says the characters are but it's to refer to the usual minute/hour/day now
 
@quartata I don't think so, now they appear more in those dramas depicting old dynasties; sān gēng (三更), however, appears in an expression meaning "late at night"
they might be used in some rural parts in China, but i'm just making this up
 
hm
 
I'm just saying there's a remote possibility in my head that they actually use it
but I don't know at all
 

« first day (2705 days earlier)      last day (2128 days later) »