nestorD

46 ms·

481.
by nestorD 7y ago
Battle for Middle Earth 1 is definitely one of my favorite RTS, I love how the constraints on buildings played into the mecanics of the game (not a fan of the second one which was more classic). By the way, if someone is still playing out t
482.
by nestorD 7y ago
That's called the multiple comparaison problem in statistics and it is both well kown and compensated for in most studies (the hard part being not to have too mauch false negatives in an effort to keep the quantity of false positives c
483.
by nestorD 7y ago
One very important thing to know about the 0.05 threshold, and which I did not found in this thread, is that the ideal p-value for a problem is a function of the number of samples (and the effect size but it as a lesser impact). 0.05 is way
484.
by nestorD 7y ago
The paper "Frequentist and Bayesian inference: A conceptual primer" has some great things to say about the p-value and why a bayesian approach might solve some of the conceptual problems : https://www.researchgate.net&#
485.
by nestorD 7y ago
It costed a lot but I have seen adverts for such hidden radio in magician's magazines from the time. There has always been a very lucrativ market for an engineer+magician that could provide hidden communication devices at the fringe of
486.
by nestorD 7y ago
It sounds like the cloud monad defined by MBrace for F# : https://lenadroid.github.io/posts/fsharp-cloud-monads-part-1...
487.
by nestorD 7y ago
As someone said on the thread, there is a clippy suggestion to tell you that you could pass by copy given the size of your struct (without it I would never have passed a struct by copy).
488.
by nestorD 7y ago
Note that you can uniquely identify all floats by printing them with printf(“%1.8e”): https://randomascii.wordpress.com/2013/02/07/float-precision...
489.
by nestorD 7y ago
Well, word2vec has been reduced down to matrix factorization with Glove (the reformulation into linear algebra managed to get similar result but one order of magnitude faster) so it does make sense to try that first.
490.
by nestorD 7y ago
I believe they recognise the problem is a multiobjective optimization problem (hence the formulation of their sentence) but their algorithm is not parametrizable : it is a single point on the pareto front and you would need other algorithms
491.
by nestorD 7y ago
I have one (but it was much more active when I was younger) and I found it very useful to go through reasonings. However mine is not critical of me, it stops at pointing logical flaws in my reasonning (as others said, it might be a matter o
492.
by nestorD 7y ago
Rust with its macro system and explicit target of system programming seem like a perfect fit.
493.
by nestorD 7y ago
Have you got a reference for katoGo ? edit: nevermind https://blog.janestreet.com/accelerating-self-play-learning-...
494.
by nestorD 7y ago
GP can take the structure of the space into account (you just need to take it into account into your mutation and crossover functions) and algorithms are typically compared on enough runs to provide statistically significant results (29 run
495.
by nestorD 7y ago
tldr: they simulated the formation of galaxies with general relativity and an alternativ theory and found a property that differs between both simulations and should be measurable with the Square Kilometer Array radiotelescope. (a nice thin
496.
by nestorD 7y ago
> They likely compared the binnary64 with posit16 using the accumulator (aka quire). Which I always find deceptiv because nothing stops us from using a quire with classical floating point arithmetic. It is in fact a comparaison betwen a
497.
by nestorD 7y ago
I am serious (and recommend reading the paper I quoted to have a good understanding of the trade-of offered by posits). Overall, posits are a new trade-of that will give you better precision (nothing exact, it is still an approximation) whe
498.
by nestorD 7y ago
Well, those will not be solved by posits, it is still an approximate floating point format. What it does is redistribution the precision to what Gustafson considers a better default and dropping edge cases in order to get more bits for prec
499.
by nestorD 7y ago
Yes, those implementation are slow compared to classical floating points. The goal, as I understand it, is to let people experiment with Posits in order to prove that they bring something to the table.
500.
by nestorD 7y ago
What if you get into the war at a later point ? Most tank with a small serial number will have been destroyed (tanks tend to have a short life) and, using the mean instead of the maximum, you will get a seriously biaised result. You could
501.
by nestorD 7y ago
F# adds a [] operator to access containers (mostly arrays and dictionnaries) and unifies arithmetic operators (no more +.). While I believe those make for a more streamlined language (I do prefer F# to Ocaml), it also means that you regular
502.
by nestorD 7y ago
Ocaml also standout as a language in which you can omit type definitions (the only exception I ever needed was when deserializing data).
503.
by nestorD 7y ago
> Do you think there's any other architecture that would've had better results -while still not taking too long to train? You could easily finetune a pretrained GPT2 model on your lovecraft dataset : https://github.c
504.
by nestorD 7y ago
It is not my domain but, as far as I know, before finding neandertal DNA in current humans we thought they probably had been exterminated but now the statu quo seem to be that they have been integrated and have mixed with our species. I bel
505.
by nestorD 7y ago
I though about it (why use Rust if your do not need high performances and can do with a GC) and the two good reasons for me would be the quality of the ecosystem (Cargo is great and you can find solid libraries for most use case) and the fa
506.
by nestorD 7y ago
It actually takes work to become good at linking pictures (which I rarely find highlighted in courses that cover mnemotechnic, probably because the author has forgotten he went through this step). It took me around a year of regular practic
507.
by nestorD 7y ago
No really, computable real store the AST of the computation to lazily compute digits as they are needed (which is an interesting idea but memory hungry at scale) while Haskell approach is more akind to a form of general type deduction. Note
508.
by nestorD 7y ago
The couple paragraphs above this citation are also worth reading, in particular: I find this very troubling. The magnesium supplementation was harmful enough to do a lot of cumulative damage over the months involved, but not so blatantly h
509.
by nestorD 7y ago
You can also use ML to learn a metamodel : a model trained on an accurate (but too costly to be run in real time) simulation.
510.
by nestorD 7y ago
To my knowledge there is no clear dominant package but things are improving quickly : https://areweguiyet.com/

More ›