nestorD

53 ms·

571.
by nestorD 8y ago
There is a very good talk on "The Unreasonable Effectiveness of Structure" that detail how just adding this structural information instead of droppping it can give great, low-hanging, rewards : https://www.youtube.com&#
572.
by nestorD 8y ago
So they took a NN taking embeddings as input, added a first layer going from one-hot encoding to embeddings (you just use the embedding matrix as first layer), did a SVD on it and retrained the network on top. Using a SVD+retraining to comp
573.
by nestorD 8y ago
In France I have seen a clear growth in the variety of products commonly available (which is great). Regarding taste it is still not clearly better than actual meat but I have tasted good supermarket-bought vegetarian burgers. In Germany&#x
574.
by nestorD 8y ago
This article discusses the Blub Paradox essay and some of its shortcomings. The original essay is one of the few texts that stuck with me. I read it not as an affirmation that some languages are strictly superior to others (in particular li
575.
by nestorD 8y ago
The linked article is not the original Blub paradox essay and does highlight this problem (the first element in the "A few problems with the BlubParadox" list).
576.
by nestorD 8y ago
In his "You and Your Research" talk, Hamming focusses on the selection of the problem you should work on as a researcher. I believe it does apply even if you are not in academics : https://www.youtube.com/watch?v=a
577.
by nestorD 8y ago
To my understanding, experiments with Unums showed shortcoming that Gustafson didn't anticipate and lead to Posits which drop the fixed length constraint. Doing that makes improving precision a lot easier but at the cost of computation
578.
by nestorD 8y ago
A Tselin automata returns always true or always false depending on its internal state (wich is basically a counter set during the learning process). Using one automata per literal per clause you can learn any logical formula in normalized f
579.
by nestorD 8y ago
They say that their results are interpretable due to being logical formulas but, later, show a toy problem in which their formula has 10000 clauses. By that definition deep-learning is probably interpretable. However I do think that it is a
580.
by nestorD 8y ago
The Tsetlin Machine being a specialized method, it makes sense that it beats neural networks in its domain of predilection (decision from, potentially noisy, binary inputs).
581.
by nestorD 8y ago
It reminds me of "A most wanted man", a movie I would strongly recommend if the subject interests you (I cannot recommend the book since I have not read it yet).
582.
by nestorD 8y ago
F# would be a good candidate. It covers all dominant paradigms without exploding complexity.
583.
by nestorD 8y ago
Herbie is not an exact match for your need it might be useful nevertheless : https://herbie.uwplse.org/
584.
by nestorD 8y ago
Note that lazy / eager is just a default : it just take more lines of code to write lazy datastructures (such as Okasaki's ones).
585.
by nestorD 8y ago
The good : - I used it as an opportuity to learn Rust and it is definitely a great way to practice with a new language. - I managed to do all but two puzzle within 24h of their publications. - There are some gems (solving the nanobots part2
586.
by nestorD 8y ago
And it might not even work. Some solutions work only on a fraction of inputs due to a vicious tie-breaking rule. This thread might help you if you already pass all the tests : https://www.reddit.com/comments/a6f100&#x2F
587.
by nestorD 8y ago
I have a hard time understanding how deepfake change the game : we could already use doubles/actors and trick footage if we wanted to forge a video. We might become more critical of videos but footage from an untrusted source has never
588.
by nestorD 8y ago
With Python instead of Fortran : "One thing that those of us in programming have certainly become aware of over the years is that people don’t mind programming with a bad language or an incomplete language. As long as they get some us
589.
by nestorD 8y ago
77MB for the Linux 3.1.3 kernel sources according to the "Implementation" section of the article.
590.
by nestorD 8y ago
I don't have a mac but it looks like incomplete documentation. Since Vs Code and dotnet core both works on Mac, you should be able to follow the linux instructions from these videos without changes : https://youtu.be/-u
591.
by nestorD 8y ago
When the first version of dotnet core went out, parts of it such as system.drawing were just empty shells. Nowadays dotnet core is all you need (happily F# seems to be developed first and foremost by the community and not by Microsoft).
592.
by nestorD 8y ago
I have been using F# for data science, developing algorithms (where ML truly shines) and most of my scripting needs for 3 years now (it is not my main language as, these days, I need to instrument some C++). As remify said F# came from Ocam
593.
by nestorD 8y ago
My ML of choice is F#, I came from OCaml because I wanted better tools, librairies and an improved syntax. The only place where I suffer from windows only librairies is UI and graphics in general (I don't do web developement so I canno
594.
by nestorD 8y ago
It was not invented yet, the "Mathematical Model for the Determination of Total Area Under Glucose Tolerance and Other Metabolic Curves" paper came out in 1994... https://fliptomato.wordpress.com/2007/03/
595.
by nestorD 8y ago
An interesting control would have been to flip a single bit but also flip a single char so that it flips several bits and compare the results. edit: Reading the linked article, they describe one of their validation method: > The Host hea
596.
by nestorD 8y ago
As a side note, I have build a C++ refactoring tool and I would have loved to be able to do it with Clasp but the lack of documentation and tutorial (probably because the project is still young) drew me away (the only thing I could find is
597.
by nestorD 8y ago
You drop denormals and change the distribution but do you keep the 0,5 ULP (round to nearest) garantee from the IEEE standard ? And are your rounding errors exact numbers in your representation (can you build Error Free Transforms) ?
598.
by nestorD 8y ago
It is a trade-of : they find improvements by losing precision where they believe it is not useful for their use case.
599.
by nestorD 8y ago
> It would be a lot harder to know what the problem is in 8 bits when everything is under question if you don't know what the outcome should be. I might have a solution for that : I work on methods to both quantify the impact of you
600.
by nestorD 8y ago
> C# is a bit nicer than Java, but you can use Kotlin or Scala on the JVM. With .Net you get F#

More ›