shoo

281 ms·

1. 2131.▲
   by shoo 14y ago

   The Turbulenz Engine is also equivalent to Thomas the Tank Engine, for some sense of equivalence.
2. 2132.▲
   by shoo 14y ago

   Regarding that first excerpt you quote, I parsed that as the author providing an obvious exaggerated example of entrepreneurial bullshit. Surely it is not intended to be serious.
3. 2133.▲
   by shoo 14y ago

   Assume that the validation data is similar to the test data. Then, by performing experiments (making submissions) and recording measurements (analysing the returned leaderboard score) you can attempt to infer something about the structure o
4. 2134.▲
   by shoo 14y ago

   I think that the Kaggle competition you describe has essentially the same structure as the Kaggle competition outlined in the blog post, and is susceptible to the same potential hazard of overfitting to the validation score. Here by validat
5. 2135.▲
   by shoo 14y ago

   Feature engineering is allowed, but some of the competition datasets are very "black box" in that very little information is given about the nature of the data.
6. 2136.▲
   by shoo 14y ago

   I feel that this is a good example of Rich Hickey's simple-vs-easy disjunction. If you are an experienced java programmer, then it is easy to understand code with a similar syntax to java. Code with a foreign syntax will be initially very d
7. 2137.▲
   by shoo 14y ago

   The author of the article does not appear to be a proponent of functional programming, or at least does not appear to be a proponent of the syntax used by a number of functional languages. These code examples are used by the author of the a
8. 2138.▲
   by shoo 14y ago

   Or with numpy: 2 * numpy.arange(1, 11) The usual notation for working with elements of vector spaces is nice when working with arrays of numbers.
9. 2139.▲
   by shoo 15y ago

   Thanks for sharing your VM & games -- I've been reading through the docs & source and finding them both interesting & inspirational.
10. 2140.▲
    by shoo 15y ago

    Do you plan on exposing parameters that control the fitting process? E.g. loss function / tree depth / min samples per leaf? Or will the fitting process always be a black-box automagic call with no user-controllable knobs? Is there any plan
11. 2141.▲
    by shoo 15y ago

    Maybe we could allow the quality of the approximation to degrade: if 'hacking is like painting' and 'hacking is like eating' are ||hacking - painting|| < \eps and ||hacking - eating|| < \eps respectively, then ||painting - eating||
12. 2142.▲
    by shoo 15y ago

    Javascript integers can only reliably represent values up to 2^53, as they are implemented by 64-bit floats [1]. This means that fib(73) and below are correct, but fib(74) and onwards are almost certainly going to be wrong. The one-milliont
13. 2143.▲
    by shoo 15y ago

    a few other tricks: - increase the cost of traversing nodes that are within known enemy firing zones - in some circumstances if the destination is unreachable it might be desirable to still compute a "best effort" path to the node that is t
14. 2144.▲
    by shoo 15y ago

    Experimental settings like these are also interesting because they provide opportunities for the application of active learning. For example, if we wish to learn the properties of some family of new materials, we may have to choose which pa
15. 2145.▲
    by shoo 16y ago

    even the images load slowly! (try going left or right...)
16. 2146.▲
    by shoo 16y ago

    And unless you were especially lucky, the existence proof would be non-constructive.
17. 2147.▲
    by shoo 16y ago

    i can view it fine in chrome & ubuntu via chrome --user-agent="Mozilla/5.0 (Windows; U; MSIE 9.0; WIndows NT 9.0; en-US))"