nestorD

47 ms·

511.
by nestorD 7y ago
It is not a full course but, since the fast.ai library does cover tabular data[1], they touch the subject in lesson 4[2]. [1]: https://docs.fast.ai/tabular.html [2]: https://course.fast.ai/videos/?lesso
512.
by nestorD 7y ago
Rust GPU iterators based on a Futhark like approach are one of my dreams (a sort of GPU equivalent for Rayon). The Rust community is used to high level functions via the iterator abstraction and wants high performance : it seems like a perf
513.
by nestorD 7y ago
Well some candidates where affiliated with the yellow vest.
514.
by nestorD 7y ago
I remember an annonce saying that the go branch of leela-zero had finaly reachen the level of facebook's network. It is arguably not Alpha zero but it shows that, if they have not reached the level of Alpha-zero yet, it is only a matte
515.
by nestorD 7y ago
Better coverage of the space of weights (less clumps and holes). If I understand the lottery ticket hypothesis [0] correctly, this would lead to a better exploration of the space and thus better results. [0] https://arxiv.org&#x2
516.
by nestorD 7y ago
I would argue that there is a need for a proper British PHP compiler.
517.
by nestorD 7y ago
And, for not much more effort (computing the variance of your samples), you can use UCB1-tuned [0] which gets rid of the 'c' parameter and tends to be even better. I personnaly think that it should replace UCB1 as a baseline when
518.
by nestorD 7y ago
She strive to have a code with no repetitions, just like their language. As such, if you try and go too far, it might become as understandable as their language.
519.
by nestorD 7y ago
Independently of the scaling, I wonder if someone tried to initialize a deep neural network with a low-discrepancy sequence (also called quasi-random numbers) instead of a uniform or gaussian distribution.
520.
by nestorD 7y ago
Here is a previous HN thread on the subject : https://news.ycombinator.com/item?id=6585828 I believe the main pain points still apply.
521.
by nestorD 7y ago
There is also the fact that intermediate operation between floats are done in float precision and not double precision (as would be the case with C). It might be the behaviour one would expect but it can lead to serious degradations of a nu
522.
by nestorD 7y ago
William Kahan, the father of IEEE floattig point arithmetic and one of the main references on numerical errors, has some bad thing to say about doing floatting point computations on the JVM : http://people.eecs.berkeley.edu/
523.
by nestorD 7y ago
Some game have more states than there are atoms in the universe. They are solvable in theory but we will need a bigger universe to put this theory in practice...
524.
by nestorD 7y ago
> Interpretability is a way of getting information. But a model does not have to be simple to provide reliable information about the relation between predictor and response variables; neither does it have to be a data model. The goal is
525.
by nestorD 7y ago
Especially what kind of usage that could not already be achieved by asking a human to write a text.
526.
by nestorD 7y ago
If I remember correctly, they say that since the training set contains extracts of question-answer sessions, it will detect the pattern and follow it when you give an appropriate prompt. So yes, you just feed the question and, detecting tha
527.
by nestorD 7y ago
I believe that, for R, releases are synonymous with distributions. The code is compiled on the user's machine (I once had a fortran error message while installing an R package). If reproducibility matters, you can also use Microsoft R
528.
by nestorD 7y ago
There is the Gosper curve : http://www.toroidalsnark.net/mkss3-pix/CalderheadJMM2014.pdf https://en.wikipedia.org/wiki/Gosper_curve
529.
by nestorD 7y ago
They also caught my eyes. I finally found a paper here (the correct denomination seems to be RAVL tree) : http://sidsen.azurewebsites.net/papers/ravl-trees-journal.pd...
530.
by nestorD 8y ago
A very good essay, also a deeply depressing one if you do not get to the end and consider his solution viable.
531.
by nestorD 8y ago
Here is a link to the unofficial html edition : https://tgvaughan.github.io/sicm/toc.html
532.
by nestorD 8y ago
For french the best equivalent that I know of would be the littré[0] but I would also be interested in alternatives. https://www.littre.org/
533.
by nestorD 8y ago
According to Wikipedia, PRISM (the collection of internet communication) is "the number one source of raw intelligence used for NSA analytic reports". https://en.wikipedia.org/wiki/PRISM_(surveillance_program)
534.
by nestorD 8y ago
F# has been open source from the start and runs on Mono but having dotnetCore on Linux gives us a faster and more reliable platform which is great. You can easily call C# code from F# so the fact that most of the .net librairies have been d
535.
by nestorD 8y ago
I translate magic books (with a speciality on old books that I find for my editor). Like others in this thread, it started as a hobby and evolved as I made contacts. It feels quitte good to know that a hobby is now earning me more money tha
536.
by nestorD 8y ago
TLDR: A reflexion on wether graduate school is a form of pyramid scheme.
537.
by nestorD 8y ago
I have read "quantum computing for the very curious" and I found it impressive, its a great article and a great way to teach. For your records, here are the few downsides I found : - Some questions seem to be of little relevance (
538.
by nestorD 8y ago
Knulp, a short novel from Hermann Hess, is a good literary take on such caracters.
539.
by nestorD 8y ago
Its a HAL website, the french equivalent of arxiv. Most french research institute have one to publish openly. https://en.wikipedia.org/wiki/Hyper_Articles_en_Ligne
540.
by nestorD 8y ago
LLVM has a wasm target so, in theory, clang should be able yo target it without modifications . In practice I found this short guide : https://stackoverflow.com/questions/45146099/how-do-i-compil...

More ›