« first day (1456 days earlier)      last day (3384 days later) » 

12:06 AM
@MartinBüttner Not sure if that's what you want, but not allowing cells to be used more than once turns it into one heck of a set cover problem
Like you could have a ton of Ks on the board, but maybe all except one need to be used by other languages and only one of them is the intended CJam one
 
boggle board now?
 
Yeah
 
hm, I think I'd prefer reuse.
that makes the designing more interesting
 
Would be better for languages which need an explicit print command imo
 
which would be better?
 
12:11 AM
Well that way the challenge isn't all about trying to find where all the "print"s are
(Which would bloat the board size)
 
yeah, I was asking which one "that way" is :D, but I think you're talking about allowing reuse
 
Ahaha oops yeah. I'm still thinking about it though
As you say it'd make designing interesting, like if you accidentally leave an "and" next to one of the programs people could have "print(20)and a" "print(20)and ab" "print(20)and abc" giving a very large number of unintended programs
But I guess that's only a problem if you say your board has, say, 50 Python snippets
 
lol yeah
 
@Sp3000 please come to the Fortnightly challenge which is on data structures.. Since it was your idea, I believe that u would know more about your own idea than anybody else...
 
Ahaha thanks :) I need to head off in a bit though, but I'll take a read through what's been said so far later tonight
 
12:28 AM
@Sp3000 ok..bye!
 
 
1 hour later…
1:30 AM
@trichoplax u here?
 
Yes here too...
 
@trichoplax I was thinking about writing a new nueral network post.. I would like your opinions...
 
Sure. Are you going to sandbox it?
 
@trichoplax it will be ready in a few minutes... basically it is similar to this: ai-junkie.com/ann/evolved/nnt4.html
 
A lot of that is about techniques which won't be available to you here.
 
1:38 AM
@trichoplax meaning?
 
No wait - I'm thinking of the wrong question - just remembered this is a new question
ignore me :)
I'm getting confused talking in two rooms...
 
1:53 AM
0
A: Sandbox for Proposed Challenges

T AbrahamMaking a neural network Challenge As you can see from the title, you are supposed to make a neural network. The task is to identify whether a given picture contains a number using neural networks. I will give a set of pictures with numbers and without numbers. The pictures will be a 8x8 grid li...

 
Will have a look now
 
@trichoplax thanks!
 
You have some left over parts from your previous wording (mentioning cats)
 
@trichoplax oh yeah.. I used the previous question to help me write this one...
 
2:05 AM
@trichoplax saved it.. what did u think?
 
I think you will find the same criticisms mostly apply to this one. You will need to give an objective way of deciding what counts as a number
 
@trichoplax what do u think would be a way to figure out what counts as a number?
 
@TAbraham It doesn't matter how it's defined, it's just important that it is unambiguous, otherwise we can't tell who wins. One answer may judge things as numbers that another answer disagrees on. There needs to be a way to decide which one wins. You can choose whatever definition you like as long as it can be applied with no doubt
 
@trichoplax u have an example definition?
 
A simple example would be "here are the pictures of the 10 digits 0 to 9. Only these exact pictures count as numbers"
 
2:12 AM
@trichoplax I thought that was implied...
 
A more flexible definition might include more than one image for each digit, to allow different ways of drawing the same number
To a human being, the number 4 with one pixel more on the tail would still count as a number 4. You need to decide whether you are asking the program to recognise any number 4, or just the particular one you happen to have drawn.
Incidentally, have you checked the license of the image you used before posting it?
 
@trichoplax why? there is no license..
 
If there is no license then you do not have permission to post it to Stack Exchange, since Stack Exchange has a license that says anyone can use what you post. That's why we are not allowed to post things unless we have the authority to share them
 
by default then, you are not allowed to use it
 
@trichoplax opps.. then what should I do?
 
2:18 AM
@TAbraham until you get permission to post it, you should delete the image
 
@trichoplax ok...
 
If you search for images with a permissive license then you will be able to post them here, provided you comply with the license (for example, some require you give attribution)
 
@trichoplax i'll try that...
 
Some search engines have an option to search images by license type
 
@trichoplax google?
 
2:21 AM
That's one example :)
 
@trichoplax how but?
 
Do you know how to search for images?
 
@trichoplax yes, not just by licenses...
 
Once you are looking at your image results, look around to see if there are any drop down boxes to further refine by things like date, size and license
 
 
6 hours later…
8:15 AM
@xnor Sorry I saw the relevant paragraph about 2 seconds after I posted the comment, oops
 
np, i do that too
 
Interesting problem though, do you know of any other approaches (if yes don't say)
 
i don't
 
Hmmm
This seems like a pretty interesting problem for Python, but I can only try when I get back home
 
9:09 AM
@Optimizer damn, I knew I should have got up and posted the CJam answer to CH's challenge instead of sleeping for another half hour :P
 
but the question is, would you have figured out the sorting thing ? :P
(in time)
anyways, I don't think the question is going viral, or my answer is getting huge upvotes
 
yes, that was the first thing I checked
 
before even realizing that I did the same thing ?
 
as in, the first thing I checked when I saw the question an hour ago ;)
 
oh, why did u not post ? slept again ?
 
9:11 AM
@Optimizer I need two more answers on the kolmogorov complexity tag though, because starting tomorrow there's are tag badges for it ;)
@Optimizer yeah, I read it on mobile and fell asleep again
 
starting tomorrow ?
 
CH's challenge is the 100th
(with the tag)
 
ohhh
yeah I remember you saying the 100 question requirement
I am 11 questions away.. though I have the upvotes.
 
I think the only tag where I'm behind in upvotes but have the number of answers is
 
... which , I already have ;)
 
9:16 AM
I know :P
 
All because of this answer
 
haha, yeah would help :P
 
9:39 AM
I think CH chose the wrong time for posting that challenge
 
it's not the kind of challenge that goes crazily viral anyway
 
still, 1 answer in 8 hr is pretty bad for CH
 
9:59 AM
added my own answer, just for the fun of it ;)
this is how I'd picture this to play out in a CJam/Prelude hybrid:
"pu htob meht knard ehs ,tunococ eht ni emil eht tup ehS"N+!
"emid a rof ti thguob eh ,tunococ a thguob rehtorB"N+!
"pu htob meht knard ehs ,tunococ eht ni emil eht tup ehS"N+!
"emil a rof ti diap ehs ,eno rehtona dah retsis siH"N+!
"pu htob meht knard ehs ,tunococ eht ni emil eht tup ehS"N+!
 
why opposite ?
where is the indentation of print here ?
 
there's no need to indent it if you can print an entire string with a single command
the ! are already in the right order
 
oh
 
and in reverse, because I'd probably implement string with a string mode like Befunge and ><>
as long as you're in string mode, each character is pushed a its character code. when you end string mode, all characters pushed are collected in a string, like CJam arrays
 
you forgot the real score in your answer
 
10:07 AM
I contemplated doing in ><> but I couldn't find a satisfyingly short enough way :/
 
@Optimizer 2045?
oh
120 permutations
 
the real score is the number of perms
 
@Sp3000 heh, I first looked at Befunge and ><> but couldn't come up with anything decent
@Optimizer fixed
 
@Sp3000 like Martin's satisfyingly short .. :P
 
Heh. Well since Martin's posted that maybe I will...
 
10:09 AM
@trichoplax lol, thanks for pointing out the type. fixed.
@Sp3000 source layout challenges are boring if people don't answer them in esoteric languages :P
I wish Marbelous was possible, but I don't think it is with a single line
 
@MartinBüttner I find most of my typos are correctly spelled incorrect words. Sometimes not just a single character out, as if the error happens early on in my brain process rather than at the physical typing stage.
 
@Optimizer now it is, I golfed it down hard!
 
Yesssss 10 bytes
 
Claps slowly
 
@trichoplax lol yeah... well I guess your muscle memory is still trained to type in certain patterns, so if you're a little off you quickly type an entire wrong word. plus, Freudian slips.
@trichoplax regarding your other comment, yes that's probably right. however I think it'll also make it more obvious where the programs are, especially if one or two are easier to find, as that would narrow down the search space significantly. I think allowing overlap will actually put more emphasis on clever board design, which I think is a good thing.
you're not the first one to bring this up though, so I'll definitely consider it ^^
 
10:16 AM
@MartinBüttner Yeah those two, plus thinking ahead to the next sentence so some of those words slip through into the current sentence...
 
I typo most of the times because both my hands go out of sync
 
@MartinBüttner I didn't mean it as a positive or negative - where you want to put the balance between cop difficulty and robber difficulty is up to you... I guess the length of programs affects this too, so it might affect which languages are advantaged
 
hm, that brings up another important point though... should program length be specified and affect the scores?
 
I am thinking of a maze based answer-dependant or CnR challenge
 
there is a CnR proposal in the sandbox (which has been there for ages, i.e. since before CnR was a thing)
I'm not sure the user is still active, so you might be able to use that as a starting point
 
10:24 AM
link ?
 
yeah he hasn't been online since September
2
A: Sandbox for Proposed Challenges

ranguDesign and Solve a Maze (this question on hold while the details are ironed out) Your task is to play the roles of both characters in this scene from Inception. In it, Cobb gives Ariadne a challenge: You have two minutes to design a maze that takes one minute to solve. Some liberties wi...

 
oh no, not an exact maze
but maze based coding challenge
 
oh okay
 
oh. KOTH flappy bird player ? :D
 
10:42 AM
Doing this in ><> is hard :( need to keep track of line numbers
 
@Optimizer how do you turn flappy bird into a multiplayer game?
@Sp3000 is that possible without g?
 
You can push an initial 0 and increment it every time you go to a new line (making sure you skip said 0 on other lines)
 
@MartinBüttner some sort of criteria on who gets to press the button
 
Twitch plays Flappy Bird
You have to plan for moves 2 games over due to the lag
 
2 games before ?
 
11:30 AM
I thought I had it but then I forgot that each program needs to print the line individually too :/
 
11:55 AM
Okay, that took too many bytes :/
 
 
1 hour later…
1:17 PM
better than prelude !
 
Ahaha true :P
 
2:10 PM
I wonder if there is a beta site that has its top user be the same as the proposer
 
If only you could write cross-site data queries
 
2:34 PM
that moment when you get a new submission to your challenge, recognise the name of the author (which is actually quite common), and it actually is a friend of yours back from school.
 
You confirmed?
 
Nice :)
Decent score too
 
why will 4 people star that message ..
 
why not?
 
2:41 PM
I mean, its not a thing to like
an equivalent FB post would be "My GF just left me because I was a jerk" - 4 people like this.
 
@Optimizer well it was a joke. that's most certainly not what happened, since he hasn't logged in to any other SE accounts either since. so I guess he's on holiday.
 
.. or he lost faith in all SE sites because of Calvin!
 
3:43 PM
Also, a star means "useful/interesting" rather than "I like this". Even bad news can be starred to bring it to the attention of the community.
(although that's not the case for everyone for this particular message - at least 25% of the starrers did so because they found the comment funny)
 
so u ?
 
lol yes
I wouldn't like it if Peter was genuinely leaving PPCG
PPCG benefits from having all sorts of incompatible parts
I see Peter as the weapons-grade duct tape that holds it all together
4
 
hm, the only difference is the scoring, right?
 
yeah, kind of ..
 
well, I'm not comfortable with dupehammering it, but feel free to VTC
oh I can't dupehammer it, it's not tagged code golf
voted to close
 
well, now that I think about the scoring criteria, this is quiet a bit different
:D
 
is it?
I don't know.
I feel like I'd just golf it, and then replace a few operations by others which use more characters but lesser presses
but algorithmically I don't think I would change anything, because it's not obvious per se what other implementation would use fewer key presses
 
4:41 PM
hmm, just that now a lot of external factors jump in ..
choose the first char of each number, avoid consecutive chars from same number
still keeping the code length not too much
 
5:24 PM
@trichoplax say, when you checked out the boggle proposal earlier, did you also see the jigsaw proposal?
 
@MartinBüttner No - I just looked at the one
Although now I have to look...
 
lol
thanks
 
0
A: Sandbox for Proposed Challenges

Martin BüttnerSolve a 2x2 Eigensystem code-golfmathcomplex-numberslinear-algebra For those with a little linear algebra background, the challenge is as simple as this: determine the eigenvalues and eigenvectors of a given complex 2x2 matrix. You may skip ahead to The Challenge for I/O details, etc. For those...

 
@MartinBüttner Was a moved to the bottom right rather than bottom left?
 
yes, thank you
 
5:29 PM
Also I took a few seconds to understand since I was initially thinking of "bottom left" as an absolute location. For some of us "down and left" might be more immediately obvious in meaning.
 
ah, good point
 
Did you type this in a mirror? The h is also right rather then left... :P
 
wow, what did I do...
 
:) lol
 
thanks a lot, hopefully all fixed now
 
5:33 PM
I take it the invalid examples are invalid because adjacent pieces must be moved as one?
 
yes. I can clarify that. in the first one df has been separated, in the second a has been removed and in the third df has been rotated.
 
Yes an explanation would be handy to speed up understanding. Maybe the offending pieces could be in bold?
Oh no it's a code block
 
you can use bold in code blocks by using HTML, but it's hardly visible for the monospaced font, so not usually worth the trouble
I added an explanation
 
I suppose it's hard to highlight the absence of the a anyway. The explanation helps a lot
 
5:47 PM
I'm guessing that scoring by number of answers gives a big advantage to people who answer early. If you don't want to just make the most recent answer the current winner, you could weight based on position so early answers are not worth as much
 
@trichoplax yes, good point
I tried similar weighting on the Meta Regex Golf, but that wasn't very well-balanced
most importantly, I'd like to include an incentive to use fewer pieces in the score.
a) to make people try harder and b) because feersum has a point that once there are enough pieces it's possible to lock the existing pieces into place by making a block that spans the width and the height ... and at that point the challenge will become pretty difficult for all but some esoteric challenges.
(so if I incentivise people to use fewer pieces it'll take longer to get there)
 
You could also add an incentive to keep the pieces disconnected
Maybe just including the number of disconnected sections in total in the score
 
hm yeah... I don't want to overcomplicate it too much either though
 
did you make sure to be explicit about whether repls, functions etc are allowed?
 
"Your program has to print N, the number of your answer, to STDOUT or closest alternative. It has to be a full program and must not assume a REPL environment."
Hm, maybe a scoring system like 10 + N - number_of_pieces_added per answer? (And scores are just added for all answers)
alternatively, 10 + N - number_of_pieces_add + number_of_disconnected_groups
 
6:03 PM
how about number of moved pieces instead of disconnect groups
 
@feersum That still allows connecting the groups with newly placed pieces
 
so?
 
The disincentive to connect groups isn't essential, it just means the challenge is likely to stay answerable for longer
But that was your point in the first place, right?
 
number of disconnected groups doesn't seem that nice as then your score would not change if you scattered around some random isolated characters
 
Ah I see - characters that don't contribute to your code?
 
6:07 PM
in fact that rule increases sabotaging possibilities
 
@trichoplax this is impossible to determine
 
you could put hcars on 10 different lines
 
I was thinking surface area too but that would have the same problem.
 
and if a language doesn't have block comments it cant be used
that will be true in general I suppose, after a while
 
@MartinBüttner yes exactly - there's an incentive to just throw in extras so ignore me...
 
6:08 PM
@feersum which rule, yours or mine?
 
but adding more pieces without reason should be discouraged
 
@feersum yes definitely
 
number_of_disconnected_groups
 
right
I like number_of_moved pieces
 
@MartinBüttner minimising that?
 
6:09 PM
10 + N - number_of_pieces_added - number_of_groups_moved
 
Would that allow a negative score for N=2?
 
yes, but I doubt that will happen
in fact, it'll always allow a negative score if people don't tend to cluster their characters
 
emo jigsawer
 
So the highest score wins?
 
@trichoplax yeah. I could either just choose the answer with the highest score, or the user with the highest score (and then accept his last or highest-scoring answer)
@feersum maybe it's worth it if you want to take a bullet for the challenge to go on :P
 
6:13 PM
I quite like the running total idea - it gives people at least some incentive not to sabotage future answers...
 
ah that's a very good point
 
Still not foolproof but I'd rather some incentive than none
I'm off for a bit.
 
thanks for the feedback!
 
You're welcome - nice to be able to return help :)
 
6:35 PM
Lady's and Gentleman....my controller : github.com/matovitch/RatLab/tree/master !! Seems to work...not thoroughly tested though. Do whatever you want with it ! :)
 
wrong chat room? ;)
 
ideed
 
wrong spellings ?
 
1
A: Sandbox for Proposed Challenges

Sp3000Fortnightly Challenge #4 - Data Structures Join us in the Fortnightly Challenge Chat to work out the details of this challenge! This is a placeholder post while we work out a list of ideas. Ideas presented so far: Submissions make up nodes in a network through which a Controller needs to sen...

 
@MartinBüttner For Jigsaw, can you put down 0 pieces?
 
6:49 PM
@Sp3000 hm, I thought I had written at least one somewhere, but I'll double-check... and now with being able to move pieces I'll actually reconsider it if I did
okay I can't find it
 
I just saw up to 10
 
I think I'll allow 0, what do you think?
actually, no, that's probably not a good idea
 
I think if the challenge is print N, and I've got enough numbers and operators on the board, rearranging's going to allow for a lot of numbers before you're forced to add more
 
yeah I was just gonna say that
 
In fact you'd probably run out of languages faster in the worst case
 
6:52 PM
okay, I'll make it at least one piece
 
it probably would score worse than adding though
 
hm, probably
(I haven't edited in the score yet)
 
I hope people don't clump all their prints together so that everyone else needs to drag it around as one unusable piece :P
 
7:40 PM
Sandbox post for Fortnightly Challenge:
 
8:05 PM
@Optimizer you can remove the first half of your comment on bacchusbeale's answer: he said the newlines aren't needed
 
I am just stating the question's core requirement
 
You can remove the first half of a comment?!?
 
obv u can edit it
 
Yes I'm just being silly :)
 
@MartinBüttner how's the challenge going?
 
8:08 PM
lab rat race? see for yourself ;) ... I can't complain :P
 
@MartinBüttner no, the data structures one..
 
I don't know
 
@MartinBüttner yeah it's good..
 
I think a lot of the answer deserve a lot more upvotes, but I guess most of them are not that interesting to look at if you're not participating in the challenge yourself.
@Optimizer I don't understand why you removed your comment completely now.
 
the code is correct
(as far as I can see)
oh well
 
8:19 PM
it's only one piece of code though, how does it print 5 different lines?
 
@MartinBüttner I want to participate, but still need to design a working bot..
-1
A: Sandbox for Proposed Challenges

T AbrahamMaking a neural network Challenge As you can see from the title, you are supposed to make a neural network. The task is to identify whether a given picture contains a number using neural networks. I will give a set of pictures with numbers and without numbers. The pictures will be a 8x8 grid li...

@MartinBüttner what do u think?
 
about what? the sandbox proposal?
 
@MartinBüttner yes..
 
honestly, I have to agree with xnor. it's still all very vague. people unfamiliar with neural networks will have no idea what to do. also, with only 10 fixed valid patterns, there doesn't seem to be any reason to involve neural networks in the first place. and it's also not clear how much training data there will be.
 
@MartinBüttner the training data will be added soon...
 
8:27 PM
posing a neural network challenge is pretty tough, and I think you should try getting some experience with simpler challenges first and also participate in some challenges yourself to get a better feeling for the community
3
 
@MartinBüttner it seems that there are some people here who know about neural networks...
 
that doesn't mean it's fine if your challenge is unclear for the others though ;)
 
@MartinBüttner what's not clear?
 
@MartinBüttner I definitely have to agree with the middle part, "there doesn't seem to be any reason to involve neural networks in the first place"
 
@KevinBrown so? there are so many projects that doesn't need to be done the way it has been specified...
 
8:31 PM
I also thought there were other "determine the number"-style questions asked before
 
@TAbraham someone without previous knowledge about neural networks will probably have no idea whatsoever what he's supposed to do. a little general background plus links for further reading would be helpful.
@KevinBrown there's one I think
18
Q: Recognize handwritten digits

aditsuYour task is to read an image containing a handwritten digit, recognize and print out the digit. Input: A 28*28 grayscale image, given as a sequence of 784 plain-text numbers from 0 to 255, separated by space. 0 means white and 255 means black. Output: The recognized digit. Scoring: I will tes...

 
I might have seen it off-site, but it was almost the same thing as this
It might have been for letters (or hex digits?) though
 
8:57 PM
@MartinBüttner i'll add links on neural networks then...
@MartinBüttner so what else do u think I need to do?
 
@TAbraham Just to reiterate what has already been said, it's always good to start off with smaller challenges (like codegolf.stackexchange.com/q/607/148 was) before moving on to the more complex ones
 
mostly what I said in my comment that got three stars. I think the challenge is too ambitious to be pulled off without a fair share of experience on PPCG. I'm not sure I could pull it off myself.
 
@trichoplax please join...
 
@TAbraham I don't have anything to add that is any different
I'd love to see a neural networks question but so far no one has made one
This isn't due to lack of interest, it's just a hard subject to write a good question for
Why don't you suggest neural networks on the Fortnightly Challenge meta post and see if people are interested? I think the one way it might work is as a collaborative project
Also, generally it's easier to write a good answer than a good question (in my opinion)
 
@trichoplax ok i agree..
bye everybody!
 
9:08 PM
yes, suggesting it as a theme might be a good idea. but only do this if you're willing to give control over the challenge to the community to some degree (as the person suggesting it, you'll naturally have your say in the challenge, but be prepared that it's going to be a collaborative effort)
wrong room
 
I'd second (third?) the suggestion that you start off with simpler things to practise with. By answering a few questions you'll get to learn about what questions you like and what can go wrong with them, which will give you more insight into what to be careful with when you come to write your own questions
 
@MartinBüttner ok.. thanks for the suggestions! Gota go! Bye!
 
9:40 PM
yay, @Peter is back ... we were getting worried over here ;)
 
9:58 PM
lol..
 
hm, I'm considering to dupehammer that range challenge (dupe of codegolf.stackexchange.com/q/38691/8478)
 
@MartinBüttner I've been travelling for a family funeral, so PPCG hasn't been top of my priorities this week. If I hadn't left my computer in sleep mode with a browser tab open on chat, you probably wouldn't have seen me until tomorrow.
 
@PeterTaylor Oh, I'm sorry to hear that. (But yeah, we figured you were travelling.)
 
 
1 hour later…
11:11 PM
@MartinBüttner (With respect to labelling trees) I think your greedy idea is based around the maximum independent set problem, which is NP-hard for a general graph. But you're given that you're operating on trees, so I would expect there to be a linear time solution. Probably too late, but...
 
@PeterTaylor Oh, I wasn't actually using it anyway, because I totally forgot about the Facebook Hacker Cup when it was on.
But they actually published solutions.
And yes, there's a linear solution.
facebook.com/notes/1047761065239794 "Corporate Gifting" if you're interested
 

« first day (1456 days earlier)      last day (3384 days later) »