eutectic

43 ms·

361.
by eutectic 10y ago
Even in C you can e.g. use singleton structs to distinguish different semantic types.
362.
by eutectic 10y ago
Being able to average over every possible computer program is so powerful that it doesn't really matter.
363.
by eutectic 10y ago
More specifically, every incorrect expert is removed at each step, and the remaining experts have their probabilities uniformly rescaled to sum to 1.
364.
by eutectic 10y ago
I wonder if there would be any point in using this to fuzz the Rust compiler. On the one hand, you could use CSmith with a C compiler as a convenient oracle, but on the other you would only be covering a very limited subset of e.g. the type
365.
by eutectic 10y ago
It's more a lamentation than a mistake.
366.
by eutectic 10y ago
The chance of Brexit was hovering around 25% just before the vote; I wouldn't describe that as 'so terrible'.
367.
by eutectic 10y ago
Degree matters.
368.
by eutectic 10y ago
0. Performance. 3. Interface documentation.
369.
by eutectic 10y ago
You could train a language model from a corpus of writing by e.g. ESL students or young children, and then disallow any words with p<p(thousandth_word) given the preceding context. :)
370.
by eutectic 10y ago
Related podcast: http://rationallyspeakingpodcast.org/show/rs-157-dr-herculan...
371.
by eutectic 10y ago
If you use Laplace succession then 50/50 corresponds to an expected success rate of 51/52 = 98%. If you use the KT estimator you get a hair over 99%.
372.
by eutectic 10y ago
If the contestants are using deep learning, I don't see why it should be any more difficult to generate a meaningful , low-dimensional representation of the game-state from raw pixels than from an abstract view input.
373.
by eutectic 10y ago
Except that Google's self-driving cars do run Linux (after all, us humans do okay with 250ms latency). https://www.linux.com/news/linux-leads-self-driving-car-move...
374.
by eutectic 11y ago
Some of my favourites: heritageradionetwork.org/series/cooking-issues/ http://rationallyspeakingpodcast.org/ http://www.microbe.tv/ (home of This Week in Virology, TWiParasitism, and TWiMicro
375.
by eutectic 11y ago
Even cryptography relies on unproven assumptions; we just consider them trustworthy enough to rely on.
376.
by eutectic 11y ago
And the cost of building a fab is increasing exponentially; eventually that trend has to come to an end.
377.
by eutectic 11y ago
If the problem is that the class probabilities change too rapidly wrt the input data, then maybe a solution would be to add a term penalizing large gradients to the cost used for backpropagation.
378.
by eutectic 11y ago
I don't think this is a new phenomenon, or that there's any reason to suspect it's getting worse. As long as charisma works people will use it, even if they have nothing useful to say.
379.
by eutectic 11y ago
They can provably tell you that a number is composite.
380.
by eutectic 11y ago
This would also hopefully solve the 'splotchiness' seen in some of the images, which presumably comes from the network optimising each pixel independently without caring about continuity.
381.
by eutectic 11y ago
[citation needed]
382.
by eutectic 11y ago
By a version of Google's DeepDream sampling from a relatively low layer.
383.
by eutectic 11y ago
Even in the West the 0.1% are much, much richer than the 1%.
384.
by eutectic 11y ago
In any other field, specifying a function by its value at a handful of points would be a bad joke. If you don't know what you want your code to do then tests will just make it harder to experiment, and if you do know then there's
385.
by eutectic 11y ago
Not sure; I copy-pasted it back from hackernews and it works fine for me.
386.
by eutectic 11y ago
And in Haskell (feel free to include this): import Control.Concurrent (threadDelay) import System.IO main :: IO ()
387.
by eutectic 11y ago
Buiding a strong evaluation function is only a small portion of what is needed to build a competitive Chess Engine. It is necessary to have a fast (and reasonably accurate) evaluation function paired with some form of tree search. Not to
388.
by eutectic 11y ago
I wonder how fast this compiles; other compilers which use extensive global analysis for optimization (like Stalin Scheme and MLton) are impractically slow.
389.
by eutectic 11y ago
> The only alternative is to implement an ad hoc interpreter. You could also have your eDSL calls act as data constructors for an AST and then implement an `eval` function which invokes a run-time compiler.
390.
by eutectic 11y ago
What do you have against parsec?

More ›