chas

34 ms·

61.
by chas 7y ago
In a Haskell context, that function’s type would be something like ‘(Num x, Num y) => a -> b -> (a, b)’ where ‘Num’[0] is the type class of types with defined numerical operators. If there is no fat arrow (=>), you can only perf
62.
by chas 7y ago
Or you aren’t doing something physical. For example there are tons of things in math that can use as much precision as you want. For a toy example, looking at rates of convergence or divergence in extremely small regions of the Mandelbrot s
63.
Computational Quadrinitarianism (2018)(comonad.com)
12 points by chas 7y ago|1 comments
64.
by chas 7y ago
I think it's important to be precise about how the monad abstraction and type system features interact in order to combine pure and impure code. I wrote a comment elsewhere in the thread ( https://news.ycombinator.com/it
65.
by chas 7y ago
Note: The approach of structuring the interactions with the IO type with the functions ( bindIO :: IO a -> (a -> IO b) -> IO b ) and ( returnIO :: a -> IO a ) is still using the abstract idea of monads to organize the impure c
66.
by chas 7y ago
Sorry for the additional pedantry, but I think this important to be precise about given the target audience of your comment. Monads aren't the separation between purely functional and stateful code. The Haskell type system maintains th
67.
by chas 7y ago
If you want to be productive in Haskell, the Monad typeclass is an important tool to familiarize yourself with. That said, unless you are working on the internals of a few libraries, you don't really ever need to know serious category
68.
by chas 7y ago
I wrote a something[0] that demonstrates Haskell programs side-by-side with a Java program for a super toy problem. It goes on to explore more complex Haskell abstractions which implement the same simple program using less approachable tech
69.
by chas 8y ago
If you want to run a computation on more than one of those AI (and scientific linear algebra) chips, you need some network to connect them. The higher the bandwidth and the lower the latency of that network, the less likely the network perf
70.
by chas 8y ago
Similar to how linear transforms can be represented as 2-dimensional arrays of numbers (that is to say matrices)[0], tensors are a higher dimensional analogue with a rich theory in their own right and a representation as higher-dimensional
71.
by chas 8y ago
That doesn’t really apply in this case though because the major thing people are using the increase in parallelism for is running larger computations or more parallel computations of the same size, rather than trying to run the same computa
72.
A Rosetta Stone for Haskell Abstractions(reduction.io)
2 points by chas 8y ago|0 comments
73.
by chas 8y ago
I think Practical Foundations for Programming Languages [0] does a good job on that front. [0] https://www.cs.cmu.edu/~rwh/pfpl/2nded.pdf
74.
by chas 8y ago
Machine learning is a bigger set of techniques than deep learning. Linear regression and random forests still work fine and can get good results in many problem domains without particularly much compute hardware. In addition non-learned fea
75.
by chas 8y ago
It makes it so all of the chips they will use have all of their neighbors, so the process (and thus electrical properties) will be more uniform between chips near the edge and chips near the center.
76.
by chas 9y ago
I don't think that aspect of centralized banking infrastructure has "very good UX" either. I could imagine something like PGP where the key management around hardware wallets and an interface like https://etherscan
77.
by chas 9y ago
Yup, all-in the sign-up is roughly the same, the bills are pretty easy to keep paid, and fees are far less inconvenient than the possibility of losing control over a financially-significant private key.
78.
by chas 9y ago
“Very good UX” doesn’t really describe interacting with public cryptocurency networks for me between the security concerns, transaction times, and obscure address names. Is there a particular use case that you have in mind where that’s an i
79.
by chas 9y ago
Here's how engineers thinking about welding joints: https://www.youtube.com/watch?v=SZiEoN8tYvI You'll note in those calculations that he assumes that the weld is basically the same as the base metal. As long as t
80.
by chas 9y ago
Here's a collection of RISC-V cores in Bluespec: https://github.com/csail-csg/riscy
81.
by chas 9y ago
Associative operators with an identity are extremely common in programming, inverses are much less common. Most ways to turn a lot of data into a little data have a monoid hidden in them with the identity serving as a default value in the c
82.
by chas 9y ago
It's useful when you write functions that take generic monoids as arguments, such as many aggregation functions on data structures. For instance, finding the largest element in a tree, or the first successful result in a list of action
83.
by chas 9y ago
I have used it recently. I have a similar attitude towards m4 as shell scripts: they are good for quick hacks, proofs of concept, and very simple systems. When I need to sprinkle just a bit of macros on something, m4 is great. For example,
84.
by chas 10y ago
Which research are you referring to?
85.
by chas 10y ago
I think that Abstract Algebra has the same relationship with CS as Linear Algebra has with the theory of most engineering disciplines. That is to say that in computer science, Abstract Algebra is the natural setting to define and decompose
86.
by chas 10y ago
This approach surprised me. Why are they doing feature extraction and then feeding that into a DNN? It seems much more straightforward to have the input of the network be noisy samples and the output be clean samples a la super resolution[0
87.
by chas 10y ago
With the exception of the antennas designed with genetic algorithms[0], the "alien" aesthetic in all of these examples looks like the output of topological optimization tools. For example: https://www.youtube.com/w
88.
by chas 10y ago
Not web development per se, but Clos Networks[0] were originally designed for telephone systems and re-emerged in large datacenter networks[1]. [0] https://en.wikipedia.org/wiki/Clos_network [1] http://resea
89.
by chas 10y ago
It is[0] and electrical engineering students make them pretty regularly, it's just much more expensive and complicated if you actually want to make a chip with the output of one instead of just simulating it. [0] https://www.
90.
by chas 10y ago
Yes. Most neural net applications separate training and inference so that once the net is in production it doesn't change, though it might be replaced with a new net rather frequently depending on how much the distribution of data it&#

More ›