nestorD

36 ms·

391.
by nestorD 6y ago
With the caveat that most movie depict people holding war bow for unrealistic amount of time, treating them like modern weapons whereas it takes non-trivial strength to hold a bow once it is armed. See this article for more information: ht
392.
by nestorD 6y ago
To give a very short summary, xoshiro256 is much faster but also has a much smaller period (however the period of the Mersenne Twister generator is way too large for most applications anyway).
393.
by nestorD 6y ago
You could run into problems with little endians and big endians if you do a binary tranfer or, more likely, have a lossy float->string->float conversion.
394.
by nestorD 6y ago
Sadly true. That's the only portable way I found to implement something akind to a Kahan sum so that no compiler / flag would destroy it.
395.
by nestorD 6y ago
For me making a habit of thinking about problems, writing my ideas and organizing them has been extremely valuable in all fields where I applied it (I am a PhD student in computeur science and act as a consultant for some magicians). I do n
396.
by nestorD 6y ago
"You should call it entropy [...] nobody knows what entropy really is, so in a debate you will always have the advantage." John Von Neumann to Claude Shannon (also why there is clearly of market for the concept of dynamic entropy:
397.
by nestorD 6y ago
I found this hypotheses in "The Blue Sense" (which gives other interesting related ideas) but it might have an even older origin. Nevertheless, it is an intriguing idea.
398.
by nestorD 6y ago
I would recommend the 1 actually. It lets you play most of the famous scenes from the movie (and I prefered it from a gameplay point of view).
399.
by nestorD 6y ago
Cardinal Mazarin's "Breviarium Politicorum Secundum Rubricas Mazarinicas" (translated from Latin to French, by Umberto Eco, as the "Bréviaire des politiciens") would be such a book (note that whether it was actually
400.
by nestorD 6y ago
Many comment here are about the first section (which is indeed rediscovering the basis of optimisation) but I find the second section much more interesting. He notes that computing the gradient is much more expensive than doing a single eva
401.
by nestorD 7y ago
What if you copy paste code that is under a viral license ?
402.
by nestorD 7y ago
Avoid reading the code, being inpired by it and using the same pattern (or worse, snippets) in your project (which could result in legal actions against you). I have often heard it in the context of windows operating system developper which
403.
by nestorD 7y ago
Note that this was true at a time (some of my math teachers learned group theory in middle school) but that nowadays math education is much less abstract until the end of highschool. At this point there is a specialized track (preparatory c
404.
by nestorD 7y ago
True, we get results good enough for that usage but not for general applications. But it is a problem well studied in machine learning and for which you can get off-the-shelf networks to use in conjunction with this paper.
405.
by nestorD 7y ago
True but their collab demo uses a separate network to infer depth information from a simple RGB picture (its a classical task nowadays) so it is not a problem in practice.
406.
by nestorD 7y ago
Not at the moment. They even compare with facebook's algorithm (the proposed new algorithm seem to do much better thanks to inpainting).
407.
by nestorD 7y ago
It relies on torch and openCV: - I have never tried running openCV explicitely on CPU but I believe it is doable. - It is trivial to run torch on CPU instead of GPU (just comment the line that sends the code to the GPU).
408.
by nestorD 7y ago
To me this seem attractive for those that want to put ML into production. Having a solid type system and an easily interoperable language could lead to a lot less bad surprises than using python. For research this seem indeed like a bad fit
409.
by nestorD 7y ago
I would love to see a cleaned-up implementation of that (and I do have notes on the subject but no time to try and implement them). Note that to be useful you need backward differentiation (a function with one output but several outputs) an
410.
by nestorD 7y ago
As Julia illustrates, you do not need to do the vectorization and GPU part within the differentiation. You can have a code using a GPU array library and just differentiate it (which ends up being more flexible / composable).
411.
by nestorD 7y ago
Nice, there are definitely people trying to push more of their datascience/ML stack to Rust and a good dataframe implementation would be useful. As a side note, a small usage example in the readme would be good.
412.
by nestorD 7y ago
Yes but its not mature enough to kill Tensorflow in the short term. I still see people prefering Tensorflow over Pytorch because they have the feeling that it is more mature for production use. Meanwhile Jax has not converged on a recommend
413.
by nestorD 7y ago
The reasonning of Tycho Brahe, taken from "The Great Ptolemaic Smackdown" (which could probably provide a much better answer to your questions) : - Procyon has the same diameter and brightness as Saturn. - If Procyon is much farth
414.
by nestorD 7y ago
On that subject, I recommend reading the wonderful "The Great Ptolemaic Smackdown"[0] blogposts. [0]: https://tofspot.blogspot.com/2013/10/the-great-ptolemaic-sma...
415.
by nestorD 7y ago
Someone else gave a link to an openai blog post, I personally first heard about it in a post by facebook (on their experiments with small precision). I believe it is not used because it requires 16bits precision which, nowadays, you only ge
416.
by nestorD 7y ago
Deep learning models require that all of their component are differentiable in order to fit their parameters. This means that most building blocks for neural networks are basic linear algebra and not much else (I am simplifying, nowadays we
417.
by nestorD 7y ago
I was mentionning manually translating a code base from one language to the other. I am not aware of a tool to do that (which, as others have said, would be tricky to do automatically for some constructs).
418.
by nestorD 7y ago
No, I am referring to proper list comprehensions (as you would get in python): https://en.wikibooks.org/wiki/F_Sharp_Programming/Lists#Usin... Its true that you can't literaly transpile very high order things
419.
by nestorD 7y ago
With the strong caveat that floating point numbers are technicaly discrete. Anecdotal consequence: 16bits floating point numbers have so much non-linearity in their round-off error that you can use them to build neural network with no activ
420.
by nestorD 7y ago
In no particular order: - significant whitespace and no `let ... in` which makes for a more streamlined syntax - no +. vs + - a solid standard library - much improved tooling (see the ionide plugin for vscode) - you can use all the .Net lib

More ›