The Nineteenth Byte - 2016-04-09 (page 6 of 16)

« first day (1897 days earlier) ← previous day next day → last day (2937 days later) »

Anonymous

6:00 AM

Tu quoque isn't a valid argument

Anonymous

Appeal to hypocrisy

Anonymous

I was referring to your earlier comment about how "other people do it too"

Anonymous

Tu quoque (/tuːˈkwoʊkwiː/; Latin for, "you also") or the appeal to hypocrisy is an informal logical fallacy that intends to discredit the validity of the opponent's logical argument by asserting the opponent's failure to act consistently in accordance with its conclusion(s). Tu quoque "argument" follows the pattern: Person A makes claim X. Person B asserts that A's actions or past claims are inconsistent with the truth of claim X. Therefore X is false. An example would be Bill: "Smoking is very unhealthy and leads to all sorts of problems. So take my advice and never start." Bill: "I'm going to...

Anonymous

Failure to always apply the rules correctly does not make the rules invalid

Anonymous

I do. You're normally a positive contributor to chat. I don't want one small incident to upset you to the point that you don't come back.

6:05 AM

@PhiNotPi I'm confused :(

Oh, - is negation, not subtraction

Well.

@QPaysTaxes Kenny cracked my program: codegolf.stackexchange.com/a/77439/2867

How are people doing D?

Anonymous

Not :P

6:22 AM

Lower bound for complexity 1 is K

What about complexity 2

@Mego accidentally killed everything by doing list() on a 2**32 len iterable, and then forgot was running pypy so couldn't figure out why pkill -9 python3 wasn't working D:

in other news, my c-large runs in < 2 sec

@Maltysen wait why were you even doing that

by mistake

Anonymous

People do silly things sometimes

@Maltysen do you still skip factors over 1m?

6:26 AM

nah i took your suggestion

the main speed up however, was something really stupid

So you're up to D?

@Maltysen ?

I was doing cartesian product (0,1) len N

and then rrejcting any that started or ended with 0

now I do len N-2 and prepend and append a 1

what i did was keep the number an integer, and turning it into a base 2 string. (also, i start at 10000......000001 and increment by 2)

what did you import (and how long was your code?)

@MarsUltor from itertools import product

and 24 lines including blank lines

20 lines without

in chars?

6:32 AM

457

not golfy at all though

you?

i was semi-golfing it (using 1-char variable names because they're faster to type), 562 chars, no imports

no imports because i think imports are overkill for this

@Mego try C-large first before trying B again

Anonymous

The thing is, my solution for B should work

Anonymous

I have no idea why it doesn't

@Mego check your formatting?

6:37 AM

@Mego Do you think that cops can provide only a(1000) and a(1001)?

but

Anonymous

@KennyLau ?

@Mego In this codegolf.stackexchange.com/questions/77419/…

Anonymous

@Maltysen It's not a formatting issue

B isn't even that hard

Anonymous

6:39 AM

I know, and that's what bothers me

did you get the small example cases?

@MarsUltor did you do B-large? can't figure it out

Hello

Anonymous

Yeah I got the small example cases easily

@Maltysen I did, using the same code as b-small

just not entirely sure my assumption is correct

if my assumption is correct, you can do it in <10 operations for each case

@MarsUltor that's crazy

Anonymous

6:43 AM

@Maltysen Could you humor me for a bit and tell me what you get for -+-+-+-+?

well, i mean 3 per char

@Mego 7

Me too

@Maltysen Is b-large too slow?

Anonymous

I'm gonna go put a hole in my wall with my face

Anonymous

I accidentally re-used a variable and it messed up the case numbering

6:45 AM

:P

@MarsUltor well i'm not making any assumptions like that and b-large will give me a graph with 2**100 nodes

@Mego livestream it

that's why I extracted it into a function

@Maltysen *up to

then ouch

user image

Anonymous

@Maltysen No thanks you people don't need to see what I look like

@Mego ^^

6:47 AM

Tip:
There are n contiguous sections
Assuming n can only decrease by 1 per flip (I don't see how it can decrease by more)
Then you just need to count the number of sections
Add 1 if the last one is a -

@MarsUltor you are a genius

> Congration you Done it

Hilarious.

Anonymous

Oh crap B-large be eating mah memory

ihaichu fractiles

should I do that distributed stuff?

@Mego So, MLE?

(Memory limit exceeded)

Or did you successfully do it?

Anonymous

6:52 AM

I got a MemoryError yeah

Anonymous

Being smarter this time

Anonymous

I erred too much on the space side of the time-space tradeoff

Anonymous

Gonna run out of time :(

@Mego for B-large?

Anonymous

Oh well I still got 30

Anonymous

6:54 AM

Yep

Anonymous

35 actually

do C-large

do the thing MarsUltor and I did, and you won't even need to do anything smart

@Maltysen What are you doing now?

trying to do D, failing, thinking about doing that distributed thing

can't figure out where to enter though

@Maltysen ?

6:56 AM

code.google.com/codejam/distributed_index.html

that button takes me to the regular code jam page

@Maltysen Uhh, the registrations may be closed, as with the regular jam.

Anonymous

Donesies

@Mego \o/

Anonymous

Score 55

Anonymous

I didn't think about setting an arbitrary bound

6:57 AM

@Mego ?

Anonymous

But yeah that's smart and makes it go way faster

> arbitrary bound

Anonymous

Check prime factors up to sqrt(n) and N, where N is some large-ish number like 10k

I would use MarsUltor's idea for B-large, but feel too guilty

@MarsUltor did you get fractiles?

Anonymous

Get a machine with like 32 GB of RAM, use the ThinGraph mixin from the networkX docs, and get cracking :P

6:59 AM

@Maltysen You still need to make sure it works though

@Maltysen Nope

Anonymous

I'm not bothering with fractiles

Stuck on it

I'm doing it for the extra points

@Mego brb hacking into the pentagon

Anonymous

32 GB of RAM isn't even that much

Just in case I get too little for the small solutions or something

Anonymous

7:00 AM

Back at my university, we had a server specifically for CS

Anonymous

It had like 512 GB RAM or something like that

Anonymous

The server admin gave seniors accounts with no resource limits for funsies

@xsot wow you got fractiles

> since you need to do well in Google Code Jam round 1 to be qualified for Distributed Code Jam

unless that's another xsot in the scoreboard

@MarsUltor oh lel

7:02 AM

hi all

can anyone compile Ton Hospel's C++ code at codegolf.stackexchange.com/a/77212/9206 ?

@Lembik gcj?

@Maltysen ppcg.

I am on ubuntu and it doesn't compile for me and I would really like to try it out

@Maltysen ?

google code jam

please :)

7:03 AM

@Mego what's your username for Code Jam?

Anonymous

@MarsUltor TheOnlyMego

Anonymous

Some poser took "Mego" 4 years ago

@MarsUltor u?

I wonder if Code Jam submissions have to work for any input, or just for the test cases.

MarsUltor

7:04 AM

@Maltysen I actually got it wrong as I submitted it too hastily. I figured out the optimal answer after a bit more pondering

Chris Jester-Young

@Dennis It has to work for the input that they give you.

@Dennis IDK, but i hardcoded the filename in my submission

Anonymous

@Dennis AFAIK just the test cases. In later rounds they run your code with the input you succeeded on to verify that your code works.

Anonymous

@Agawa001 That's not very nice.

@Agawa001 ?! I am not sure I have ever even thought about CG!

7:05 AM

@Maltysen how are you taking the input?

@Agawa001 is this google code jam?

what is google jam ?

@MarsUltor for some taking filename from sysargs, others from < and >

Anonymous

Google Code Jam, a multi-round programming competition run by Google

@Lembik I think @Agawa001 means you have 0 answers in PPCG.SE

Anonymous

7:06 AM

I used to hardcode filenames, now I just use sys.argv[1] for simplicity

So hardcoding is allowed?

Anonymous

I also used to use C++, back in the dark days before I knew Python

@MarsUltor for coinjam I hardcoded N and J

Anonymous

Yep

@MarsUltor oh I see.

Anonymous

7:07 AM

Hardcoding the filename, that is

you can even do it by hand

Anonymous

I dunno if hardcoding N and J is allowed for C

Chris Jester-Young

@Maltysen You get a chance to do the distributed jam if you survive the regular jam up to round 3 or so.

f=open('C-large.in')
l=f.readlines()[1:][0].split(' ')
f.close()

Anonymous

If nobody from PPCG makes it to the finals I'm gonna be sad

7:08 AM

since I have it open anyway

^^ Not that hard to modify

@Mego oh plz dont accuse me trollin again

hi @TonHospel

Chris Jester-Young

@Mego I always use stdin/stdout and never open files directly.

@MarsUltor use with !

@TonHospel Your first code fix seems to have got rid of an error which is good

Chris Jester-Young

7:08 AM

31 secs ago, by Chris Jester-Young

@Mego I always use stdin/stdout and never open files directly.

no with needed. ;-)

Anonymous

I use something like with open(sys.argv[1]) as f: T, *data = f.read().splitlines()

@TonHospel I now get bpaste.net/show/042da8ee4d0a

@Mego what's the advantage of using with?

@Lembik auto closes and handles exceptions, etc.

Anonymous

@Lembik It automatically closes the file when it goes out of scope

Anonymous

:28883172 Syntax is with fn() as var

7:09 AM

Wait

you use the first line?

Chris Jester-Young

2 mins ago, by Chris Jester-Young

@Mego I always use stdin/stdout and never open files directly.

So, something like: ./A.py < A-large.in > A-large.out

@ChrisJester-Young Sweet! One of the problems just got trivial.

Chris Jester-Young

@Dennis :-D

thanks

@Dennis Which one?

7:11 AM

Not saying. :P

C?

in any case.. could anyone, please, attempt to compile that code :)

@Lembik ?

it is driving me mad

@Maltysen what is in the "handles exceptions, etc." part of your answer

@MarsUltor sorry..wrong person

@Lembik: Try to qualify conv with ntl: NTL::conv<...>(...)

7:13 AM

@Lembik i'm not sure. Its called contexts

@TonHospel ZZ sum = conv<ZZ>(0); becomes NTL::conv<ZZ>(0); ?

@Dennis It's probably C. In that case, it's still a lot easier to actually do it IMO

Is it too late to get involved with Code Jam?

@lembik: Yes

Chris Jester-Young

@Mego Poor thing. :'(

7:14 AM

@PhiNotPi No, ~18h remaining

@TonHospel what was the " ntl:" part of your comment?

test.cpp:100:26: error: expected primary-expression before ‘>’ token
ZZ sum = NTL::conv<ZZ>(0);

I clearly did that wrong somehow...

@Lembik: No, that notation should have worked. How about: ZZ sum; NTL::conv(sum, 0); (on two different lines so we know which line errors come from)

@Dennis How is making it work just for the test cases making it easier?

If it works for the test cases, how wouldn't it work for everything else?

Plus, you're gonna have to generate the output somehow anyway

@MarsUltor TBH, I've seen previous Code Jam problems, and they aren't really that interesting to me.

@MarsUltor Sorry, really not saying.

7:21 AM

@PhiNotPi Yeah, but many people aren't doing it because the problems are interesting

doesn't know how to start for D

test.cpp: In function ‘int main(int, const char**)’:
test.cpp:104:39: error: no matching function for call to ‘conv(NTL::ZZ&)’

@TonHospel ^^

Well, I'm going to bed now.

Bye!

@xsot how did you do D?

Chris Jester-Young

@MarsUltor Read the commentary at the bottom of the question. It explains the logic quite well.

But

7:26 AM

@ChrisJester-Young yeah but K goes up to 100

it doesn't explain how to do it without listing ecerything

K^C goes up to 10^18

@TonHospel I am looking at the conv part of shoup.net/ntl/doc/tour-ex1.html and also stackoverflow.com/a/27377278/2179021

are either of those helpful?

I don't really know how to code in C++

Chris Jester-Young

I do. :shrug:

@MarsUltor There's a very short linear time solution. I don't want to spoil it for you, but consider drawing out the tiles and carefully observe how the gold tiles propagate in each stage

@ChrisJester-Young great! :)

7:30 AM

@lembik: They mainly say all versions should work :-) However try the implicit one: ZZ sum; sum = 0;

test.cpp:103:39: error: no matching function for call to ‘conv(NTL::ZZ&)’
double rho = sqrt(conv<double>(sum))/factor;

:)

I think that means success for the implicit one

@TonHospel I saw this in the changelog "Made the conversion interface more complete and uniform. Also, using template notation, one can and should now write conv<ZZ>(a) instead of to_ZZ(a) (for backward compatibility, all the old names to_XXX are still there, but many new conversions are not available under these old names). "

Chris Jester-Young

.oO(I wonder what the advantage to using NTL is compared to Boost.Multiprecision, because I have experience with the latter.)

@lembik Yes. So maybe for this one we can try implicit too for the rho: double rho = sum; rho = sqrt(rho) / factor;

@lembik: Otherwise double rho = sqrt(to_double(sum)) /factor;

test.cpp:103:18: error: cannot convert ‘NTL::ZZ’ to ‘double’ in initialization
double rho = sum;

@lembik: split it up again ? double rho; rho = sum;

7:36 AM

test.cpp: In function ‘int main(int, const char**)’:
test.cpp:104:9: error: cannot convert ‘NTL::ZZ’ to ‘double’ in assignment
rho = sum;
^

how about using to_ZZ ?

@lembik: This is getting silly. Can you paste the file /usr/include/NTL/ZZ.h somewhere so I can see what is available

@TonHospel paste.ubuntu.com/15704659

@zyabin101 :) NSA?

@zyabin101 is that 100% polite by the way?

@Lembik Sorry, now deleted.

np

@lembik: double rho = sqrt(to_double(sum))/factor;

7:43 AM

no errors!!

@Lembik \o/

@TonHospel great new output messages :) Out of interest, does your code rely on the min eigenvalue being at least 1?

@Lembik: No

@TonHospel what happens if you try the example from the paper with high eccentricity?

@lembik: Should work for any positive definite matrix

@lembik: Works fine but no Siegel reduction yet

@lembik: So needs 109 points

7:46 AM

@TonHospel arxiv.org/pdf/nlin/0206009.pdf (36)

@TonHospel did I tell you I tried github.com/abelfunctions/abelfunctions which I think just implements that paper and it is amazingly slow

@lembik: They must have made an error :-)

@TonHospel well.. github.com/abelfunctions/abelfunctions/issues/121

@MarsUltor @Mego and other code jam players: what's your username on the code jam?

@TonHospel and the mathematica version is also very slow according to Martin B

TheOnlyMego, Maltysen, xsot, MarsUltor, DennisMitchell

7:50 AM

hi @MartinBüttner

@lembik: I also optimized the ellipsoid walker quite a bit over what they suggest (directly combined with the calculation avoiding all matrix multiplication)

@TonHospel it would be awesome if you could write this up in any case.. even in ascii in your answer

@lembik: Anyways, the runtime grows very much with requested precision and dimension, so I hope these people didn't ask for insanely precise results

@MarsUltor How can I find someone on the very large leaderboard?

There isn't any search tool, is it?

@zyabin101 top right corner

You can als add to friends

7:53 AM

@MarsUltor The last one is totally not me. >_>

@MarsUltor Uhh, I'm not logged in.

@TonHospel that could well be it.. I will test all these things :) How confident are you that your method gives the right answer now? Are there provable bounds?

nods

Chris Jester-Young

@MarsUltor OTOH, my username on Google Code Jam is easy to figure out. ;-)

totally removes DennisMitchell from friends list

7:55 AM

@TonHospel timing your lovely code now

@lembik: Yes, it uses the bound from the paper. In fact it's too pessimistice and typically returns way more precission

hahaha

"ANGEL" SONG BY TAHER SHAH

@xsot you're in the lead of our pack! (Or are you?)

@ChrisJester-Young Not for me

7:56 AM

@lembik: Especially in higher dimensions this makes the program unnecessarily slow. I might replcace it by an adaptive expansion of the ellipsoid instead (but losing the guaranteed error bound)

@MarsUltor cky.

@TonHospel This is too awesome for words!

it's not that easy though :(

@zyabin101 err my actual score is 45 unfortunately

Malyisen's is easier

7:57 AM

@TonHospel m = 10 , n = 20 takes only 30 seconds for 5 runs!

@xsot Then which username is the correct one?

I need to change the challenge now as you have broken it :)

I mean, that's me on the leaderboard, but the displayed score assumes that all large outputs are right

@lembik: And if the error bound was better it would in fact take less than 1 second

Chris Jester-Young

@zyabin101 Yep. /cc @MarsUltor

7:58 AM

@TonHospel great!!

@TonHospel I am surprised you only quoted n= 19 or 20 though.. my computer is normally slower than everyone elses

@ChrisJester-Young D: you've only done A

@lembik: In this version you can give the error bound as an optional commandline argument

Chris Jester-Young

@MarsUltor That's because I'm going to sleep on the rest of the problems.

But

@lembik: I typically work on an ancient core2 duo laptop

7:59 AM

@TonHospel oh.. I didn't realise

B and C shouldn't be that hard

« first day (1897 days earlier) ← previous day next day → last day (2937 days later) »

Transcript for

Apr '169

The Nineteenth Byte

The Nineteenth Byte: General discussion for codegolf.stackexc...

join this room
about this room

00:00

06:00

12:00

18:00

all times are UTC

site design / logo © 2024 Stack Exchange Inc; legal

mobile