Always

I'm not hearing good things about GCJ :(

Singing the song of angry robots?

Amazing, big thumbs up!

Do I?

ONE WEIRD BUG COST HIM 1.3M POINTS IN GOOGLE CONTEST. SEE WHY

(you're welcome Errichto)

Looks very nice!

I disagree -- only by 4 points :P

With bardek and DamianS we have:

A -- 17

B -- 5,822,900

C -- 5,690,448

D -- 5,028,530

E -- 5,215,303

F -- 5,340,296

total -- 27,097,494

Did you install Hightail using snap? I think that is currently not working, so try instead to just download the .jar file from the Releases page and run it by double-clicking.

It's easier to think about an axis-aligned square.

Let X be the weight that this method yields. Think about what Kruskal's algorithm would do if the weight of this edge were set to less than X, and what it would to if it were set to more than X.

Thanks. These should have been swapped.

Unfrozen now.

Could you also link to sunset's problem?

It's over. Thanks everyone! The editorial is up.

That's something that's traditionally done in ACM ICPC type contests — the standings are frozen for the last hour of the contest.

Yes.

See here

This contest is not rated. In general, there is some system actually, though I'm not sure whether it has been used recently.

Thanks! Would never have figured it out...

Is it no longer possible to see other contestants' codes in the new interface?

Or censor the stream ;)

But what's the point of doing both? In my head, Hightail replaces freopen...

Very cool!

One question: consider a problem from a 5h ACM contest; if I have a rating 2500 and the problem has 2500, does it mean that I (my team?) would have had a probability 0.5 to solve it during a "typical" 2h individual CF contest, or during a "typical" 5h ACM team contest?

Hightail now has over 6000 downloads. Thanks everyone!

New release v0.9.7

Big change: now supports Competitive Companion (can parse tasks/contests from browser, supports Google Code Jam, Facebook Hacker Cup, etc.). See video.

Also:

• snapcraft support (auto-updates under Linux)
• some fixes

Get it here :)

Note that it runs like: r14 /Users/.../a rather than /Users/.../r14 a. So you should make sure that r14 is in your PATH, for example. Try running r14 /Users/.../a from the console and see if that works.

You would need to make a script that both compiles and runs your file (like make a && ./a on unix). Then have the prepend command be g14 (with a space after) and the path from ... be %P. That should work...

Another name is Lagrangean relaxation (if I understand correctly what you're referring to).

Actually I think that's a different trick. What we do to handle 100009 is to randomly set each number to either x or x-100009. Then it's in [-100009,100009], and half of the multiplied coefficients become negative, so that it cancels out to a way smaller number (around sqrt(n) times smaller, I guess).

Hmm, I myself don't understand the numerical stability of FFT too well. Experimentaly, all implementations that we had could handle 1009. With the neat trick you're describing, they could even handle 100009 AFAIR.

Your formula doesn't display, so not sure what it says, but the modulo was only 1009, so at least standard FFT works well enough, with a maximum of like 200000 * 1009 * 1009.

It does not say "1 <= ki <= n".

The editorial is up: editorial

We apologize for the technical problems with CodeForces. The queue is now moving again. The contest is extended by 30 minutes. It will remain unrated ;)

It's fixed now.

Thanks for noticing. It will be fixed as soon as possible.

See here and here. Thanks for your feedback!

Well, I like these two: 414B - Mashmokh and ACM, 414C - Mashmokh and Reverse Operation

Wow, thanks! Didn't expect to see such a post... I guess at ICPC you learn to pick a good three-person team!

The video of today's talk by Laszlo is here, for those who are interested.

Just point it to your executable file, whatever its name is. It has some comments in the configuration dialog which should make it clear.

No, it's cross-platform.

Should work behind non-authenticated proxies. See: https://github.com/dj3500/hightail/issues/60

Well, both work for me... Does the CodeBlocks .exe work when executed from outside of CodeBlocks, like from a console?

Previous company name is ISIS, how to list on CV?

My screencast here (99th place :C)

I'm guessing div1-E was inspired by Bathroom Stalls from this year's GCJ Qualification Round?

To give an "update" on the cited post, the FHC people are making some efforts to get 2015 prizes (and some 2014 prizes, I think) paid out, but seem to be hitting some sort of wall internally (with the finance team or whoever).

So I agree: it's anyone's bet... :)

Should work for both, I would say. Whatever a logged-out user can see. (In case of AtCoder live contests, also what a logged-in user can see, if you give your login credentials.)

Well, at most ~30 seconds after the scheduled time something should happen :) If it doesn't work for you, could you make a bug report (issue) on GitHub (with details, like which contest and what you are entering in the fields)?

See dotorya's comment above.

Isn't this just maximum-likelihood estimation, essentially? Since the samples are drawn independently, the probability of the sample vector is the product of their individual probabilities. As rng_58 said, it is too small, so we take the logarithm. This is then called log-likelihood. Comes up all the time in statistics and in machine learning.

This thing (MLE) gets even weirder when dealing with continuous distributions. There, the probability of having sampled any particular value (real number) is zero, but we take the probability density function at that value instead and it still somehow works... ;)

Yes, this is actually described in the editorial. Some people prefer to think of this as a min-weight bipartite matching instance (and the Hungarian algorithm also builds matched-vertex-wise increasing solutions).

That's a great problem, so that's nice that more people got to experience solving it

Our thoughts exactly :)

We estimated that not many such people would participate, and we were right (but also: one of the reasons this was unrated). On the other hand, some people seem to have had prewritten solutions to B (and some of them got RE due to too small arrays in their codes :P). We did anticipate that the problem would be known in theory, because it is a classical question, but didn't know that it was posed as a contest problem.

See the editorial for an answer :)

Or run your MCMF algorithm for k iterations.

Besides, congrats for squeezing this in the last minute :)

Great question! Maybe one of the two teams who have done so (never red 27411549, ★SweeT DiscoverY★ 27412140) could explain their solutions?

Auto comment: topic has been updated by dj3500 (previous revision, new revision, compare).

The editorial link has been added to the first post.

Yes (at least with 'an' optimal solution :) ).

Yeah, I didn't get it either.

For what it's worth, I cannot see you on the registrants list either...

Very nice problems. Thanks DEGwer!

Seems to work now.

Good point, we'll fix that.

By the way, the writer is currently flying.

On a plane, or...?

Could someone please paste their implementation of D (which they think is elegant)?

Awesome functionality!

Now, when I try to edit an old group, I get error 500. But I hope it's fixed soon :)

Cool! Points 1 and 2 sound exciting to me. Especially 1, I guess. As far as I know, this implies having a browser extension or something like that?

I think now it all finally makes sense: Acrush: the boyband of girls winning hearts in China

Or paste the contest link before the contest (you can also use the link to the list of registrants), set the time and press Schedule.

Yes. What is your application, by the way?

My screencast

Hmm, I was doing recursive DFS in D, so it probably would have failed if there was a stack limit.

Something else that I didn't like though was getting TLE in D, which I fixed using FastIO :)

New release v0.9.6, with new features:

• added support for AtCoder live contests (requires login; thanks shanto86!)
• added support for Open Kattis (thanks ykalchevskiy!)
• option to use full names of problems on CodeForces as filenames (thanks jfsanin!)
• fixed CodeChef parsing (thanks Shinora007!)

Get it here :)

And it's done by fabianun :)

My screencast

Very nice problems, as usual!

Right. You should only count triples which are in different branches of the tree going out of L. (Note that there may be Ω(n) of these branches, so you need to do that in a non-slow way.)

That's not what the problem is asking. It asks to compute the number of such triplets.

You can prove that for any such triplet, there must exist another vertex L which is between them in the tree (i.e., L-A, L-B, L-C forms a star-like shape), and the distances L-A, L-B, L-C must be equal. So you loop over every vertex L, then run DFS counting vertices at each distance. Gives O(n^2) runtime.

• Is suitable for a general audience and does not contain any unsubstantiated, false and/or misleading claims.

Then better not write "P = NP" in a comment :)

Same here. Their prize payouts have been a joke for some time now. For the 2015 prize, the last thing I got was a tax form for businesses to fill out, like, 17 months ago.

No.

Well, we already do that, except we treat "A" as the name. This would not be too hard to implement, so feel free to open an issue and maybe someone (or you) will do it :)

Could you raise this issue on GitHub? Try to give an example of an input and a program for which this behaviour appears.

For the second one, please also open an issue. Although the chances that someone will implement this are not so high :) I always envisioned Hightail rather as a tool for situations where you solve a problem in one shot (without restarting your computer etc.).

You may be confusing bits of two or three separate occasions here :)

2011/12: Because the next ICPC was to be in Warsaw, the Polish Collegiate Contest (AMPPZ) was granted one slot, which would be awarded to the best university from AMPPZ which does not qualify from CERC. We got a good (4th) position at AMPPZ, losing only to University of Warsaw teams, so only two things could make us not go to World Finals: University of Warsaw not qualifying from CERC (highly unlikely, especially seeing that top3 on AMPPZ was all Warsaw) or a different team from our university beating us on CERC. At CERC, we performed really poorly and did get beaten by another team from our university, but their position was also not enough to qualify — there were 3 slots and they were the 4th-best university. So we were happy about going to the WF. However, then an an extra (4th) slot for CERC was announced. There was some interpretation-of-the-rules thing at that point, but in the end the other team from Wrocław got awarded the WF spot, and our spot from AMPPZ was awarded to the next university (Technical University of Poznań). Not nice, but reasonable.

2012/13: We got one place behind the four WF slots (even though we felt we performed very well, many CERC teams seem to have done an excellent job that day). We heard unofficially that they were considering adding one more slot. Then, after about one month, we did indeed get the invitation :)

2013/14: Again we got the fifth university-place on CERC. But at the closing ceremony it was announced that there will be five slots. Great :) Then, over the next month, we did not get the invitation, became gradually more worried, then learned unofficially that CERC would get four slots in the end (even though in the previous year it was 4+1, and the region was doing crazy well on WF). Two months after the regional, we appealed in an effective way and pretty soon the ICPC made it right (but in the meantime other people involved also tried to appeal on our behalf and were unsuccessful).

So my impression is that there is not much ill will, but the process should be made fully transparent. The ICPC does use some criteria that they talk about (medals from the region at the previous World Finals, for instance, and some perceived level of the regionals) and a lot of criteria about which they do not. And the Asian regionals frankly sound like a horror story to me, and possibly highly corrupt. I would say that every year, before the first regional begins, some formal description of the rules should be made available to everyone (maybe even an algorithm) which, given the results of all regionals, will determine the World Finals advancers without any further human input.

This does not sound like a pipe dream to me, because the ICPC is in fact becoming more and more open and modern (in 2011 or 2012 they have started to release tests, for example, and did you know that they are now releasing this stuff as well, or this?). Maybe in 2-3 years...

Three CodeForces bugs (two of them very old), documented in my screencast:

• the obsolete warning about not using %lld: link,
• the hacking code displayer behaving insane: link,
• problem not locking successfully: link (this one is new, I think).

You know he's not actually losing points on this, right? People should actually think of being hacked as a favour / hint :)

Except of course if someone generates a testcase against the particular primes you're using in your hashing solution :P

It's average-case constant (that is, amortized and with high probability). It's based on a hash-map — you might want to read up on these :)

My screencast (got 42nd place)

Makes sense, because problems are visible publicly. You could always stay logged out until you wanted to submit.

Actually, no — you only get rated if you submit something.

My screencast

Any news?

New release v0.9.5, with new features:

• added support for CodeForces Gym
• added support for AtCoder (but only past contests)
• for CodeForces, now you can paste the contestRegistrants link
• contests will now parse even if some problems fail to
• cosmetic changes & bugfixes

Get it here :)

Finally fixed in the new release :)

Thanks for the question! Indeed, it turns out that some testcases had no solution. This is corrected now. We have rejudged all the failed solutions from during the contest and fortunately, only two contestants were affected. The practice-mode submissions were not rejudged — feel free to resubmit your code.

Probably it should... although the registration form maxes out at 1990 :)