bravura

161 ms·

1. 1651.▲
   by bravura 18y ago

   Rejection sampling is perhaps not a good option, because the assumption does not hold: "the maximum weight isn't too much greater than the average weight" In learning algorithms, the distribution of the weights can get piqued.
2. 1652.▲
   by bravura 18y ago

   I am trying to speed up a machine learning algorithm. There are a very large number of possible output classes. It would be too expensive on each training step to perform an update for all outputs. I use an exponential moving average to kee
3. 1653.▲
   by bravura 18y ago

   Do these convertible debt-style seed-funders typically take a board seat?
4. 1654.▲
   Ask HN: Fast sampling and update of weighted items? (data structure like red-black trees?)

   21 points by bravura 18y ago|26 comments
5. 1655.▲
   by bravura 18y ago

   If I am interested taking pre-orders on selling an item, but don't want to go through the hoops of a merchant account, what is the other options for accepting payment?
6. 1656.▲
   by bravura 18y ago

   My understanding is that cPickle can be slow, because of introspection.
7. 1657.▲
   Fast deserialization in Python(blog.metaoptimize.com)

   19 points by bravura 18y ago|3 comments
8. 1658.▲
   by bravura 18y ago

   This is more than 11%. It's the FIRST link
9. 1659.▲
   by bravura 18y ago

   I'm sorry, but usually it's the opposite for me. I especially appreciate social technologies before they jumped the shark. Email? It was awesome before everyone and their mother used it. Friendster? It was like the wild wild west, free and
10. 1660.▲
    by bravura 18y ago

    I still don't get Twitter. I'm sorry, I've tried. I know that it's considered hopelessly square to be behind the twitter trend, but I just don't appreciate it. Maybe it's my phone device? Do most people use twitter desktop software? What do
11. 1661.▲
    Ask HN: Best practices for rapidly changing database schema?

    4 points by bravura 18y ago|1 comments
12. 1662.▲
    by bravura 18y ago

    Voltaire says that behind every fortune, there is a crime. But I believe the opposite. Assholes get very far, but rarely do they make it to the top. All the top people I know are actually great people.
13. 1663.▲
    by bravura 18y ago

    Or, 255 versions of it, including the original PDF: http://scholar.google.ca/scholar?hl=en&lr=&cluster=7...
14. 1664.▲
    by bravura 18y ago

    Does it support receiving SMS messages? I looked at the features, and I don't see this functionality.
15. 1665.▲
    Ask HN: How to make a test suite?

    2 points by bravura 18y ago|2 comments
16. 1666.▲
    by bravura 18y ago

    Peter, any chance you could organize them, or tag them in some way? Otherwise, there is no way to know what any data set is until you click the link.
17. 1667.▲
    by bravura 18y ago

    Here are some other dataset aggregators: http://theinfo.org/ http://infochimps.org/datasets http://ckan.org [Comprehensive Knowledge Archive Network] http://www.trustlet.org/wiki/Repositories_of_datasets http://www.dan
18. 1668.▲
    by bravura 18y ago

    I heartily recommend writing your own encryption algorithm, for the same reason that I recommend performing your own surgery. If you think you're good at something, then why should you let only people with life-long expertise do it?
19. 1669.▲
    by bravura 18y ago

    I second the nomination of Bishop. It is the standard text. It is only two years old, and Bishop will teach you machine learning the way that the field practices it nowadays. In my lab of fourteen people, we must have six or so copies of Bi
20. 1670.▲
    Ask HN: Full-text search of Unix files?

    6 points by bravura 18y ago|3 comments
21. 1671.▲
    by bravura 18y ago

    Also, what do you mean by "broad survey crawls"?
22. 1672.▲
    by bravura 18y ago

    Could you outline how to configure it to collect only textual content?
23. 1673.▲
    by bravura 18y ago

    If I'm not mistaken, wikia does not want you to download their crawls using a robot: http://soap.grub.org/arcs/robots.txt
24. 1674.▲
    by bravura 18y ago

    Write tests first: http://c2.com/cgi/wiki?CodeUnitTestFirst It's more fun that way. Instead of your code breaking when you add the test, it's broken until you add the functionality. My code has many many tests. My co-workers would laugh a
25. 1675.▲
    by bravura 18y ago

    gojomo, I have looked at the file formats. Could you propose some off-the-shelf web spiders? I would like to accumulate a lot of text for NLP research.
26. 1676.▲
    by bravura 18y ago

    Damn, how did you hear about what I'm working on? Look for an email from me within the next two weeks. You'll know it from the subject line....