imurray

22 ms·

241.
by imurray 15y ago
I've lost count of how many times I've seen statements about users being products on here. Can we please stop reciting platitudes? Obviously Google would care if they annoyed enough users to lose ad revenue, or tarnish their reputation. The
242.
by imurray 15y ago
An update to a popular submission a year ago: http://news.ycombinator.com/item?id=1449117
243.
New free stonesoup eCookbook - simple 5 ingredient recipes(thestonesoup.com)
2 points by imurray 15y ago|1 comments
244.
by imurray 15y ago
A recent critical review of that book: http://radfordneal.wordpress.com/2011/06/18/two-textbooks-on...
245.
by imurray 15y ago
Inexplicably dead comment, copied below: dawson 1 hour ago | link [dead] Me too :( If you haven't already, checkout The Haiku Project http://haiku-os.org/ (inspired by BeOS).
246.
by imurray 15y ago
The earliest example I've seen of 2π is in a 1763 letter from Thomas Bayes (the paper that appeared in the Royal Society proceedings directly after the one that's famous). He used c for the circumference of a circle whose radius is unity.
247.
by imurray 15y ago
"What Tau Sounds Like" http://www.youtube.com/watch?v=3174T-3-59Q (Yes, we know that Tau doesn't really sound like anything, but this was fun and better than I expected.)
248.
by imurray 15y ago
Searching reveals lots of previous discussion on Benford's law on here, so I won't give all the links. Of course, it's an interesting observation, so it's worth advertising every so often. Here are some hacker-newsers testing files in the
249.
by imurray 15y ago
An unrelated tip for people that use TeX and want good PDFs. If you have an old .ps file with bitmap fonts from dvips, and can't recompile the .tex source, it's still possible to get a nice PDF with Type 1 fonts: pkfix-helper input.ps
250.
by imurray 15y ago
I just found pdfsizeopt.py, which doesn't seem to be that well known(?). It's reduced the size of most of my PDFs. pdfsizeopt.py --do-unify-fonts=false --do-unify-pages=false infile.pdf The first option is important : several of m
251.
Pdfsizeopt: PDF file size optimizer(code.google.com)
2 points by imurray 15y ago|2 comments
252.
by imurray 15y ago
Yes. I'm afraid I'm heading off and can't think of a good way to hint Right Now. (It's Exercise 4.13 in http://www.inference.phy.cam.ac.uk/mackay/itila/book.html ... and there is a solution and other related weighing problems in there.)
253.
by imurray 15y ago
Fun fact: it's possible to solve this puzzle even if you have to declare which balls you want weighed against each other in each of the three experiments before you get any results in.
254.
by imurray 15y ago
The record shows a rather more reflective attitude: http://news.ycombinator.com/item?id=1796989
255.
by imurray 15y ago
If considering using fountain codes, I would definitely want to check out the IP issues; there are certainly some patents in that space. For the application being discussed, fountain codes wouldn't be an exact fit anyway. I pointed to them
256.
by imurray 15y ago
eru makes a really good point, and I suddenly feel much less confident. My intention was to sketch the idea for a protocol that was reliable (with very high probability, just like anything in the real world). So, to repeat the argument: i
257.
by imurray 15y ago
I did. Thanks! Their arguments suggest UDP with no acknowledgements. If they had needed to send more data they might have needed slightly more sophisticated error correcting codes than repetition codes. Doing that and keeping the latency lo
258.
by imurray 15y ago
I agree Luby's codes came after Quake. Although other codes for the erasure channel did exist, and the important ideas for the class of scheme I'm talking about were known. That wasn't meant to be my point though. I was just thinking about
259.
by imurray 15y ago
I liked the discussion of transmitting game state for network play. Quick summary: Doom sent differences of state, which had to be received and therefore acknowledged. Quake sent the whole game state each time, but compressed, so it would
260.
by imurray 15y ago
Except that the doom example in the paper (figure 11) shows that down-sampled images aren't recovered well. The posted work is designed for pixel art.
261.
by imurray 15y ago
Earlier comments: http://news.ycombinator.com/item?id=2601347 http://news.ycombinator.com/item?id=2578706 Some of the previous discussion isn't great (squabbling about dupes, not being able to download the pdf). A common comment was "ha
262.
by imurray 15y ago
[ldng: put some whitespace before code on hacker news: it indents it and the asterisks don't disappear.] I infer that ldng has a 64 bit machine. On my 32 bit machine, Python 3.1 is faster than 2.6 for these examples. On a 64 bit machine I g
263.
by imurray 15y ago
Yes. Given the string so far, see how probable it is that the generator would generate the next character, p(x_n | x_<n). Running through the whole string you can build up the log probability of the whole string: log p(x) = \sum_n log p
264.
by imurray 15y ago
If you have a good generator for text, you have a useful language model that can be plugged into applications such as speech recognition, OCR, predictive text entry systems and compression.
265.
by imurray 15y ago
When times are that short you should question if Python is actually doing anything, or just promising to do it later (remember that range is a generator in python3). $ python3.1 -m timeit "map(lambda x: x**3, range(10000))" 1000000 lo
266.
by imurray 15y ago
Always worth checking. Although the relative merits change when doing many joins to build up a single long string. This recommendation is to go with .join() by default: In performance sensitive parts of the library, the ''.join() form shou
267.
by imurray 15y ago
Request: could submitters of long videos add a few words in a comment to say what it's about and for whom it will be worth watching? From listening to the audio while doing other things: This talk was largely about shotgun assembly of genet
268.
by imurray 15y ago
John gives good motivation. Usually one should just use a featureful environment/library that provides this stuff. However, as he says, you sometimes want simple standalone functions. Motivations include simple distribution, or encouraging
269.
Stand-alone scientific code(johndcook.com)
48 points by imurray 15y ago|3 comments
270.
by imurray 15y ago
I do this, but warn that it sometimes leads to sites breaking: text spilling out of boxes or getting obscured by running under other elements. Web-designers: please make your sites robust to people picking their own fonts, and also robust t

More ›