stellalo

8 ms·

61.
by stellalo 4y ago
I guess you could attach the probability as weight to the event outcome, instead of taking a uniform mean. Not that without probabilities it’s “impossible”, it’s just a different metric
62.
by stellalo 4y ago
They are all univariate models: some are trained offline on a bunch of different series before being applied (deep learning, “global” models), others are applied directly to each series to forecast (“statistical”, “local” models), but the t
63.
by stellalo 4y ago
The history section mentions the work of Baidu’s founder as somewhat related to PageRank in that it scores based on links. Work of others are also mentioned in the same section as “influential” for the development of PageRank. If we just tr
64.
by stellalo 4y ago
Can anyone recommend any read on what the story of cookie banners is, why they are there, and so on? It’s the single most annoying thing in today’s browsing, and I wish it would just go. Really, is anybody glad the thing is there? Why? Aski
65.
by stellalo 4y ago
And this just happened to happen in September 2022?
66.
by stellalo 4y ago
Also a reason: you landed that job you wanted
67.
by stellalo 4y ago
> Research shows that attending faith-based services 4 times per month will add 4 to 14 years of life expectancy. I’m wondering what confounder is behind this
68.
by stellalo 4y ago
> I saw the pain points of git What are these? Asking for real: it’s the second time I read a similar sentence on HN this week, without finding any specifics, so I’m curious
69.
by stellalo 4y ago
It seems to me like the article is advocating for a solution to a problem that isn’t there? > But all technology evolves, and git is already showing its weaknesses. Such as?
70.
by stellalo 4y ago
Let’s be serious: I yet have to see a convincing argument for 0-based being better than 1-based, or the other way around. And designing a language based on “what everybody else does” is definitely not the right approach.
71.
by stellalo 4y ago
I would have thought that “what = context” and “how = code”
72.
by stellalo 5y ago
I first learned about Julia somewhere in 2013, I thought “damn, this looks really nice”, but never used it seriously until 2016 (v0.4). This language gives me joy, I never get tired of it.
73.
by stellalo 5y ago
“Perano” really sounds like it could be the name of some Italian village... oh wait: https://en.m.wikipedia.org/wiki/Perano
74.
by stellalo 5y ago
“Science is pretty clear” -> doesn’t cite any science.
75.
by stellalo 5y ago
It’s interesting to see the implementation of a priority queue, but the title is misleading: the author is speeding up a slow implementation of Dijkstra by a factor 2700
76.
by stellalo 5y ago
This is interesting, but it’s a pity that the only experimental results do not show fast asymptotic convergence of the main and simpler algorithm proposed (Algorithm 1, the one stated in the abstract) but only for its line search variants (
77.
by stellalo 5y ago
Interestingly, this happens ~17 years after Facebook went online, same as what happened with Google/Alphabet in 2015
78.
by stellalo 5y ago
> 175 billion words They probably meant “parameters”, although then the sentence makes little sense with that.
79.
by stellalo 5y ago
Well, to be honest I still don’t understand the difference: they seem to me like two different styles of structuring the same kind of logic. In fact, the linked list example in the referenced article from Mike Innes I don’t find very convin
80.
by stellalo 5y ago
Nice article! I agree with many of the points. If I know the Julia community well enough, criticism like this leads to constructive discussion and eventual improvement. One section I don’t get is that about the iteration protocol. It seems
81.
by stellalo 6y ago
That’s interesting: did you use Flux?
82.
by stellalo 6y ago
Looks like the links to the book are broken in the article. The book in question is this: https://en.m.wikipedia.org/wiki/Hacker%27s_Delight
83.
by stellalo 6y ago
This article starts off with the following premise: “one of Julia’s original promises was to take over other languages in terms of popularity” (paraphrased). Honestly, I cannot find such promise in the original announcement: https:/&#
84.
by stellalo 6y ago
> What’s the unique selling point? It’s multiple dispatch, how did the almost-author-of-a-book-on-the-language not get that?
85.
by stellalo 6y ago
Object orientation is no “must have” for a general purpose language to become popular. OO is a paradigm like many others, and it is one that in my opinion is much more abused than, say, functional features. In a few years we will look back
86.
by stellalo 6y ago
Actually, it looks like in the paper something else is going on other than subgradient steps: there is some more randomization going on, that can prevent some steps from being taken. So yeah, there is a connection with online subgradient, b
87.
by stellalo 6y ago
Recasting quantile estimation as an optimization problem is trivial: the q-quantile minimizes the “pinball” loss (see first eqn in http://statweb.stanford.edu/~owen/courses/305a/lec18.pdf ) with parameter q. W
88.
by stellalo 6y ago
That is essentially an online subgradient descent, with fixed stepsize 1, for minimizing the L1 loss.
89.
by stellalo 6y ago
Sure, so what
90.
by stellalo 6y ago
Because it’s beautiful: https://www.flickr.com/photos/lastingimages/17342493661

More ›