imurray

26 ms·

1. 211.▲
   by imurray 15y ago

   Some comments are on the original blog posting: http://www.johndcook.com/blog/2012/03/09/monkeying-with-baye...
2. 212.▲
   by imurray 15y ago

   I've been hit by nasty slow-downs with denormal numbers a couple of times. It's much easier to hit these problems with single-precision floating point, where near underflow happens much more quickly than with doubles. As a folk-theorem of
3. 213.▲
   by imurray 15y ago

   I first heard about de Bruijn sequences [1] in a talk by Persi Diaconis. Among other things, he described the lock picking "use" listed first in the Wikipedia article. They're such obviously cool objects, that I feel like they should be use
4. 214.▲
   by imurray 15y ago

   Ubuntu's current (11.10 oneiric) menu cleverness breaks things for me using a non-standard window manager. (The gnome-terminal menu appears even when set to hidden.) Solution: purge every package with appmenu in the name and reboot. I hope
5. 215.▲
   by imurray 15y ago

   The two biggest differences I noticed when I tried XeTeX was 1) the pdf files were much smaller than from pdflatex, and 2) I couldn't use microtype [1] with XeTeX. I stuck with pdflatex to use microtype (which is awesome), and use pdfsize
6. 216.▲
   by imurray 15y ago

   That is pedantic, and not in a particularly well justified way: http://public.wsu.edu/~brians/errors/nonerrors.html#between
7. 217.▲
   by imurray 15y ago

   There's a very good review of what it takes to build a state-of-the-art AI Scrabble player in the following paper (appears to be open-access, apologies if not): World-championship-caliber Scrabble, Brian Sheppard, Artificial Intelligence ,
8. 218.▲
   by imurray 15y ago

   Thanks. It works in Ubuntu (Chrome or Firefox). Playing with: http://ipinfo.info/netrenderer/index.php it seems that Internet Explorer won't display the strikeout characters because the font on HN is set to Verdana, which doesn't have the
9. 219.▲
   by imurray 15y ago

   Not in Ubuntu 11.10: % apt-cache show update-motd ...Description-en: superceded by pam_motd in libpam-modules ...The functionality formerly provided by this package is now integrated into pam_motd, in libpam-modules.
10. 220.▲
    by imurray 15y ago

    I regularly have to recompile a couple of Ubuntu packages after excising undocumented junk. I really should submit a patch with a c̶o̶n̶f̶i̶g̶ ̶f̶i̶l̶e̶ g̶c̶o̶n̶f̶ dconf option that does the same. But for now I haven't found the time and te
11. 221.▲
    by imurray 15y ago

    [dead] comment by cdibona: "Please don't sync yet, it's currently in a mixed state. The 'repo for-all git push' is still running and will take some time to complete, so if you sync now you'll get some parts with Gingerbread and some parts w
12. 222.▲
    by imurray 15y ago

    Great! Thanks. I now see that git also has a built-in --color-words option, and I've added that to my notes. (I don't have a git setup to quickly try whether that would still work in conjunction with an external diff tool like oodiff though
13. 223.▲
    by imurray 15y ago

    I'm not sure if odt2txt puts paragraphs on one line, or flows them. But either way, if you change a single word in a long paragraph you can get large-looking diffs. For version-controlled documents it's good to use a word-based diff rather
14. 224.▲
    by imurray 15y ago

    There are models that do learn several latent features for each player [e.g. 1]. And I recently supervised a Master's project motivated by exactly your observation[2]. Models with multiple features per player do make better predictions, but
15. 225.▲
    by imurray 15y ago

    as the requestor, prepare sample letter for the writer A slightly less presumptuous version of this strategy is to provide a bullet list of things they may like to include in a letter. When I'm writing letters I usually ask students for s
16. 226.▲
    by imurray 15y ago

    The statement about spinning is too weak. You have to spin rather than flip to try to reveal bias: http://www.stat.columbia.edu/~gelman/research/published/dice... As for the statistical significance, here's another view of how to do it:
17. 227.▲
    by imurray 15y ago

    One solution I heard has a basic idea that's quick to explain, although the details are messy. It might be interesting to get some intution: Flip the coin N times and count the number of heads, K. You got one of the C = {N \choose K} = N!/(
18. 228.▲
    by imurray 15y ago

    Reminds me of zip, zip3, zip4, etc. in Haskell.
19. 229.▲
    by imurray 15y ago

    http://ycombinator.com/newsguidelines.html Please: "Be civil.... When disagreeing, please reply to the argument instead of calling names." There is something in Laurie's comment. You could have focussed on why people enjoy Sudoku anyway
20. 230.▲
    by imurray 15y ago

    It would be nice if Microsoft provided a non-silverlight interface to the Feynman lectures that they have the rights to: http://blogs.msdn.com/b/escience/archive/2009/07/15/project-...
21. 231.▲
    by imurray 15y ago

    There's no need for seq or jot. In bash you can replace `seq 1 10` with {1..10}. Or {01..10} if you want zero padding when using $i to label files. I use an function called "up" that takes an optional integer argument to go up a certain num
22. 232.▲
    by imurray 15y ago

    http://mathlesstraveled.com/2011/06/04/the-collatz-conjectur... (At the end of the comments you'll see that the author of the proof withdrew the claim.)
23. 233.▲
    by imurray 15y ago

    Here is a related project on turning an image of a Go board into a computer readable file understood by most GUIs for playing Go: http://www.inference.phy.cam.ac.uk/cjb/image2sgf.html
24. 234.▲
    by imurray 15y ago

    Can anyone get the original text? It is a bit annoying how the BBC non-transparently correct their articles. The old version is cached here: http://www.webcitation.org/60evsCAry
25. 235.▲
    by imurray 15y ago

    and the geeks will enter yy/mm/dd More precisely, I think many geeks will enter YYYY-MM-DD — http://www.cl.cam.ac.uk/~mgk25/iso-time.html — having lived on two continents, I use this format whenever I can.
26. 236.▲
    by imurray 15y ago

    Separate to the whole GPU issue: when timing code in Matlab it's best to use a more careful routine than tic/toc, like timeit.m (by a Mathworks employee): http://www.mathworks.com/matlabcentral/fileexchange/18798
27. 237.▲
    by imurray 15y ago

    Thanks, I had confused myself. You're right: shifting with a negative value (or far too big) gives undefined behaviour in C. (The type punning stuff is formally undefined too, a memcpy would be more portable, but gcc promises to make the
28. 238.▲
    by imurray 15y ago

    A nasty solution, which assumes and abuses IEEE floating point format and demonstrates a couple of things. int isPowerOfTwo (unsigned int x) { int exponent; union { unsigned int u; float f; } tmp; tmp.f = x;
29. 239.▲
    by imurray 15y ago

    I completely agree with all of the above. I was doing a linear scan, and I know that that is artificial. Your stride suggestion does slow things down dramatically; thanks for the simple-to-implement idea. (Shrinking the lookup table by 1/8,
30. 240.▲
    by imurray 15y ago

    For interest rather than practical use: Creating a 2GiB lookup table is ~10% faster on my machine than method #10. That is, faster on the benchmark in the blog post. A 2GiB lookup table is horrible, has a little setup cost dominated by call