« first day (12 days earlier)      last day (4413 days later) » 

6:53 AM
24hours with no new question. TIME FOR PUBLIC BETA.
2
@Gilles I changed it to 3CNF. I hope we meant the same (:
 
 
1 hour later…
8:11 AM
@RanG @Raphael and this is why you don't use abbreviations
 
@Gilles KNF was just a slip of mine (konjunktive vs conjunctive)
Good morning to you too! :)
 
@Raphael I'm leaving for work in 5 minutes, I expect to approve an edit from you that expands the abbreviation before then
 
no way!
I told you, I don't know how to describe 3-SAT accurately and shortly w/o using CNF as abbreviation.
All proposals I have seen so far failed either one or the other
Not that I like SO too much, but look there:
jQuery: a cross-browser JavaScript library that provides abstractions for DOM traversal, event handling, animation, and AJAX
other excerpts explain abbreviations but are either too long or have the room.
(because the abbreviated thing is the subject of the tag)
maybe we should not have the tag after all? I thought it was too localized earlier. 3SAT is central,though.
cstheory.SE does have some very specific tags, though.
anyway,afk
 
8:35 AM
1
Q: Saving on array initialization

AryabhataNote: This is a seeding question, so please feel free to be pedantic etc. I will be deleting this paragraph some time after accepting an answer. I recently read that it is possible to have arrays which need not be initialized: i.e. it is possible to use them without having to spend any time tryi...

 
8:49 AM
@Gilles I tried to solve the problem by not solving it.
 
 
3 hours later…
11:45 AM
Hi @DaveClarke
 
1
Q: Unsupervised learning of Sound Files (The Joe Walsh Problem)

Dave ClarkeThe Eagles are a rock supergroup from the 70s and 80s, responsible for such classics as Hotel California. They have two quite distinctive sounds, one where guitarist Joe Walsh is present (for example, in Life in the Fast Lane) and one where he is absent. The latter songs have a markedly more somb...

 
12:11 PM
Hi
 
I seem to remember a kid doing something along what you ask for in 494 during last year's Google science fair
 
Can you get him/her to answer the question?
Because I really only want to listen to the songs with Joe Walsh.
 
:)
 
Although I'm interested in the answer, I'm mainly just seeding the site. We'll now attract Eagles fans.
When it goes public, that is.
 
Good thinking. :D Wonder when my father will end up there
 
12:34 PM
Why not incorporate that link into an answer ... perhaps after waiting a while to see what other people come up with....
 
I am not sure he published anything on that
 
That doesn't matter.
It's on the interwebs.
 
But if you can not get the algorithms or implementations, what use is it to you?
 
Perhaps I have no clue how to approach the problem. The page you point to has loads of ideas. That's surely better than nothing.
 
1
Q: Are generational garbage collectors inherently cache-friendly?

GillesA typical generational garbage collector keeps recently allocated data in a separate memory region. In typical programs, a lot of data is short-lived, so collecting young garbage (a minor GC cycle) frequently and collecting old garbage infrequently is a good compromise between memory overhead and...

 
12:48 PM
@DaveClarke That's right. I guess some kind of signal analysis is the way to go; I know next to nothing in that area.
Btw, your question titles "unsupervised". Certainly, some amount of training is acceptable?
 
Maybe someone else will be able to help me.
Let's see how we go.
 
Churning out some tag excerpts; hopefully they make sense.
@DaveClarke Certainly. All it needs are some pattern recognition guys
 
1:18 PM
@Raphael An excerpt that provides only a formal definition isn't helpful
I'm afraid if I keep rejecting your excerpts you'll be banned from proposing tag wikis for a week
 
@Gilles Which are you referring to?
 
@Raphael trees was especially bad
That's a definition of the word
It's not a definition of the tag
 
I was under the impression that the two should ideally be indistinguishable
 
@Raphael no, absolutely not
 
how would you describe the tree tag?
 
1:22 PM
Read the blog post (like it says in the reject message)
 
You never get any notification of rejections...
Hm. We do 4. almost never. 2. explicitly says "Avoid generically defining the concept behind a tag, unless it is highly specialized." -- in the context of CS, trees are a highly specialised concept.
Anyway, we are arguing too much about inconsequential things.
I would like for some voting to take place on tag wiki (excerpts). As it is, you can easily get edit wars
@Gilles You can always accept and improve afterwards. Too bad there is no improve button as for posts...
 
@Raphael there is voting: it takes two (or 3 on SO, or 1 user with 20k)
 
@Gilles I see.
That is not transparent in the interface: you only ever see suggestions that you can act on.
 
10
Q: Allow 'improve' for Tag Wiki Edit Suggestions

Lance RobertsI've been approving or rejecting Tag Wiki Excerpt Edit Suggestions, but some of them are close and I'd like to just tweak them. We can improve regular edit suggestions, how about giving us an Improve button for Tag Wiki Edit Suggestions. Note: If I click on the tag wiki excerpt link, then it ta...

 
Aw, too bad. Thanks for the reference.
 
1:29 PM
If there's nothing to say about a tag, I prefer to leave the tag wiki empty
If anyone on the site seeing the tag knows what it means, the tag doesn't particularly need an excerpt. If there's no content in the tag wiki, leave the excerpt blank.
 
I was thinking we should provide excerpts on principle and add wiki entries later
 
Oh, and the #1 guideline for tag wikis is do not copy-paste from Wikipedia
 
although I don't quite get them. In many cases, tag wikis are bound to duplicate information found on Wikipedia (maybe even of the same name)
^^
 
@Raphael better leave the excerpt blank, this way we can see easily which wikis need filling
 
But then we don't have excerpts to guide tagging (esp for non-experts) until somebody gets around to write wiki, which may be a while
 
1:33 PM
@Raphael On such diverse sites as Stack Overflow, Unix and Linux and Science Fiction and Fantasy I've found that WP content is not what we want in 90% of the times. Also, it's illegal without attribution.
 
Hello there
 
cf Jeff: "That’s why the first two pages of tags should have excellent tag wiki excerpts at a minimum. If they have great, complete tag wikis, that’s even better, but you have to crawl before you can walk. Focusing on the ~500 character excerpt is a simple way to get started" (in the article you linked)
 
@Raphael you can start small
 
@AlextenBrink Hiho
 
My god, what a week
 
1:34 PM
@Raphael yeah, but you don't need to have all that in the private beta. Let the tags grow a little first, so we have a better idea of what they mean
 
@Gilles Sure, but I can comfortably write a dozen excerpts in half an hour, wiki's take much longer. So I think as long as many tags do not even have an excerpt, leaving wikis for later is better.
 
For example your excerpt for does not describe the use of the tag (I asked about this on meta)
@Raphael that's because your excerpts aren't very good
 
@Gilles Good point. Although in many cases we already know, because they relate to specific concepts/fields in CS.
@Gilles Maybe, but also in your opinion.
 
Take the time to write a proper wiki, and extract the salient points and summarize them in 1-2 sentences. You'll get a better excerpt.
 
@Gilles Agreed. Still, the time issue remains.
 
1:37 PM
We could use more (diverse) questions before the tags can really start representing their topics
 
(oh, my current rep is my year of birth. Nice.)
@AlextenBrink I can not imagine that any amount of questions changes the meaning of or .
 
0
Q: The scheduling tag

GillesThe tag scheduling currently has 4 questions, all of which are about schedulers in multitasking operating systems. While this is technically a special case of scheduling algorithms, topics in operating systems and optimization tend to be disjoint. Should we use the same tag for both, or do we wan...

 
and are less clear, though.
 
True enough for 3-sat
We really need to get into public beta
 
@AlextenBrink True that.
The stall is built, the menu printed, ingredients are here and we are waiting for customers.
 
1:48 PM
Let's see if I can improve my array interleaving algorithm description
 
Our progress is to be discussed.
in The Assembly, 5 mins ago, by Grace Note
@Raphael The plans are actually about the be discussed, now that we're approaching your second week of private beta.
but they don't have an ETA on when we will hear from them :-/
I can not really estimate how "good" our beta is as I do not have participated in any other.
I think we have had good questions, we certainly have experts in a variety of subfields and we have motivated people. Our only weak spot is that we have not yet been tried extensively by the practitioner onslaught.
 
Which is exactly why we need a public beta - there's only so many questions a group of experts can entertain themselves with
 
@AlextenBrink Precisely
 
I just hope the powers that be aren't so short-sighted to only look at the fact that we seem to have 'run out of steam'
 
2:05 PM
Dito. Given the wind we had to sail against earlier that is not an unfounded fear, I am afraid. "Look, we told you, this is never gonna work!" But let us have trust.
I have more questions taken from exercises or exams, but there have been some complaints.
 
If I look at this comment from Shog9 (meta.cs.stackexchange.com/questions/78/…) then at least they are willing to be persuaded by reason - this gives me hope that even if the initial judgement is not the desired outcome, then at least we can express our position and maybe sway their opinion
Hmm, I think I may have an interesting question - I formulated and answered it myself a few years ago
It's about priority queues with constant-time deletion and log n insertion :)
 
@AlextenBrink Deletion of any element?
 
Deletion of the min-element :P
So insert $O(\log n)$, get-min $O(1)$, extract-min $O(1)$
 
worst-case?
 
Amortized is good enough
I hope there's not some trivial structure that caches extracts or something :P
 
2:19 PM
balanced search tree with a pointer to the smallest element?
 
And how do you delete from a balanced search tree in constant time?
That was my initial reaction as well :)
 
you might be able to do stuff with heaps by performing deletions lazily
you delete the minimum and point to the next small element, nothing more to do (except maybe rebalancing, but that is O(1))
 
Rebalancing is usually O(\log n)
Unless I'm mistaken :P
 
@AlextenBrink right, duh.
 
This may end up being quite a nice question then :)
 
2:33 PM
Definitely
For average case, skip or jump lists come to mind
 
Isn't deletion log n for those as well?
 
um
 
1
Q: Does there exist a priority queue with $O(1)$ extracts?

Alex ten BrinkThere are a great many data structures that implement the priority-queue interface: Insert: insert an element into the structure Get-Min: return the smallest element in the structure Extract-Min: remove the smallest element in the structure These structures are also commonly known as (min)hea...

 
It's worst-case log n at least, not entirely sure about average case
 
Even if you only ever remove the first element?
 
2:37 PM
You may need to remove a node in each of the log n levels
That may indeed be constant time expected yes
 
Yay, trusted user for a day! er Now I can edit tag wiki excerpts without @Gilles sabotaging me! ;)
 
Gratz :P
 
Thanks.
Are priority queues really called heaps? IIRC heaps are merely the most often used implementation.
 
That's quite possible yes
 
never seen it that way. *shrug*
 
2:41 PM
Fix't
 
I retagged accordingly, plus I switches your Os to \cal{O} ;)
 
2:54 PM
Hmm, the question has 2 upvotes and 1 view... :P
 
@AlextenBrink Caching, caching, caching. That's the answer to nearly all odd behavior like that around here.
 
Yeah, you're right
Ugh, my array interleaving algorithm is rather hard to express in code because of all the bit-level things you have to do
 
3:11 PM
@AlextenBrink That suggests it is not a very good algorithm ;)
 
I prefer the term 'involved' :P
The main issue is knowing which index is next to be swapped, the rest is easy
@Raphael I just saw that @IvoFlipse has a question about k-way mergesort - considering that I spent 8 hours per evening over the past 4-5 days debugging and implementing k-way mergesort, I might be of assistance :)
 
good lord, is it that hard?
 
Well, I had some handicaps
We test on a PC with 20 MB of memory, and we sort 1 GB of data
We have to log in on that PC with SSH
We use mmap in our algorithm, which doesn't work on Windows (which is what I run), so I can't test or even compile the code I write
We also had to do permuting and other stuff, plus we really wanted to get it fast so we tried a lot of things
And it doesn't help if you have bugs that only appear if n >= 30000000...
 
3:32 PM
@AlextenBrink Auch...
 
So, in short, the algorithm isn't too bad, but making it fast over SSH is :P
 
Well I'm going to work on Algorithms-class somewhere this week again, so when I encounter the problem again, I'll be sure to ask a question
 
Ok :)
 
That's the problem with trying to have naturally occurring questions, you don't just go out there and ask them :(
 
You mean you can't force a question to naturally occur?
 
3:37 PM
Well you first have to study something to know what you don't know ;)
 
Indeed
@Raphael I added pseudo-code to my answer as you requested: cs.stackexchange.com/questions/332/…
It ended up being not too bad at all, but a proof that it actually works will be painful...
 
Ouch @AlextenBrink.
@AlextenBrink Cool, checking it out.
@AlextenBrink haha... :P
you know that you can make Markdown use monospaced fonts inline by `test`? (becomes test)
That could make your second paragraph more readable
 
3:53 PM
Nope, let's have a look
 
also, I'd prefer B[0,N/2] or B[0..N/2] over B_{0, N/2}
Math-aware syntax highlighting, that would be cool (and probably impossible to implement)
 
There, that's better
Thanks for the suggestions :)
 
my pleasure
 
 
1 hour later…
5:12 PM
@AlextenBrink I can do O(1) expected time min-delete using jump lists, and O(1) wc-time using skip lists if you allow array copy in O(1).
 
5:47 PM
An array of $omega(1)$ elements I presume? :P
 
of course :>
 
That doesn't sound like a reasonable assumption then I fear :P
 
aww
 
But glad you like the question :D
I can give you a hint if you want, on what direction you should look if you want to get to my solution
 
6
Q: What is coinduction?

Dave ClarkeI've heard of (structural) induction. It allows you to build up finite structures from smaller ones and gives you proof principles for reasoning about such structures. The idea is clear enough. But what about coinduction? How does it work? How can one say anything conclusive about an infinite...

 
5:51 PM
not yet
 
Ok ;)
 
I'm off for now; Swedish movie night!
 
6:08 PM
Ok, have fun :)
 
 
3 hours later…
9:28 PM
Hahaha @DaveClarke, my fingers where itching to make this edit. Glad to know I am not alone. Sambo says we are all crazy. Oops.
 
 
1 hour later…
10:43 PM
@DaveClarke why on 525?
 
@Raphael why not?
(still trying to work out a semi-decent answer)
oh, you've answered
I'd grown used to the push notifications already
 
@Gilles As I understand the question, it does not ask about a proof technique.
Not sure wether my answer is fully (or even remotely) correct; my introduction in coinduction was sketchy and I have not gotten around to reading more on it.
 
@Raphael “induction (…) gives you proof principles for reasoning about [some finite] structures. (…) What about coinduction (…) infinite structure?”
@Raphael I haven't verified all the equations, but the gist of it is correct: smallest vs largest fixpoint
 
Ok; I only know coinductive definitions; coinduction as proof technique is beyond me. As you typically want properties to hold at any point in a trace (e.g.), normal induction should cut it.
 
@Raphael mumble bisimilarity mumble
(currently trying to turn mumble into English)
 
10:55 PM
I learned the word "bisimilarity" only two weeks ago while attending a workshop. It means two systems can not be distinguished from the outside, i.e. their internal actions are hidden,right?
 
@Raphael right
 
I love 494, btw. I envision a music player where "random" moves smoothly through the genres and does not jump from Amy McDonald to Kalmah to Corrs to Metallica...
 
you test the equivalence of two systems by testing that they match under all observations
 
@Gilles What's got it to do with coinduction, then? ?(
 
which entails finding a way to make the states equivalent
@Raphael the equivalence is defined from the way the systems can be observed
an inductive equivalence would be defined from the way the systems are built
 
11:09 PM
hm
 
have you studied process calculi yet?
 
not really.
I can read it,kind of, but I have little intuition and knowledge of results.
 
Is that like the pi calculus?
 
I know LTL if that helps
 
@AlextenBrink for example
 
11:13 PM
A lot of my courses use some form of process calculus then
 
concurrent systems tend to have all kinds of different observations, because they're nondeterministic
 
mCRL2 was apparently developed in one of our departments, not sure how well-known that toolset is
 
sometimes you're interested in all possible traces. Sometimes only whether certain traces are possible. Sometimes you don't care about “internal” traces
 
Had to use it last semester
 
all of this is defined from outside, P=Q means that you can't distinguish P and Q under a certain observation
most equivalences are coinductive properties of the processes
@AlextenBrink never heard of it
 
11:16 PM
Good, I didn't really like the tool :P
 
@Gilles the set of all traces is defined by the system,and the rest by deleting homomorphisms,right?
I think I see that you can get into trouble if you only consider finite prefixes; you never know when two traces are "at the same point".
 
@Raphael <thinks> yes, I think
 
11:49 PM
@Raphael, @AlextenBrink, @Gilles: Looks like you guys are learning something coming up with answers to these questions, which is great. @AlextenBrink: mCRL2 is a really useful tool. I'm using on for a paper I'm currently working on. It should be more well-known. The GUI is horrible, but the tools are super-robust and efficient.
 
The tool is useful, but the GUI really could use some work indeed :S
 
@DaveClarke Indeed!
Research SW with crappy GUIs suck. This is why we build ones with great GUI!
wwwagak.informatik.uni-kl.de/research/JAguc plug Our benchmark is that our local biologists can work with it and are happy
 
@AlextenBrink could make a lot of people happy by volunteering to fix the mCRL2 GUI as a summer project ;)
 
(well, CLI tool chains are worse than crappy GUI,I guess?
Gogo Alex! :)
 
Just adding 1 line of code fixes the obnoxious error message you get every time you close a tab on Windows for instance...
 
11:54 PM
CLI tool chains can be easier to use, once you've worked out what the chain is. Certainly, if you've used the mCRL2 tool, you'd start to love the CLI approach ... if you write a script.
 
I might, but I have a lot of other stuff I want to do as well, and not enough time to do it all :S
 
Oh, yes that's me. Why do you think I need to filter out the boring Eagles songs?
 
:D
the "Baron of Techno" worries about the Eagles--nice.
 
I don't even know what a DJ does, except put records on. Anyone can do that, surely.
 
11:59 PM
but the right ones at the right time
 

« first day (12 days earlier)      last day (4413 days later) »