LolWolf

29 ms·

301.
by LolWolf 8y ago
I don't know if it's at all embarrassing (though there are some HNers that definitely did not at least skim the paper, which is a whole other thing). The examples they gave in the paper were all tiny: MNIST was the biggest with 78
302.
by LolWolf 8y ago
“Elementary” doesn’t mean “what you learned in elementary school,” although it is publically perceived as such. Almost universally, in almost every text I’ve read, it means “use tools that are well-known to people with an undergraduate degr
303.
by LolWolf 8y ago
It’s possible to do this subsets of stuff, but rewriting QM in a better way is at least as difficult as rewriting the linear algebra of hermitian operators in a way that is easily palatable. While QM is somewhat of a new development, linear
304.
by LolWolf 8y ago
I just read you paper, it’s quite fun! This comment was made with the belief that you claimed that all infinite graphs have this property. Anyways, I know it’s an old thread, but if anyone is reading this, I highly recommend reading it. Her
305.
by LolWolf 8y ago
Another fun one (from the many examples above!) is a new paper I am working on at the moment as well, but in the intersection of physics and math: it turns out you can phrase a bunch of physics problems as optimization problems, then use so
306.
by LolWolf 8y ago
Is this true? I mean, consider the infinite directed graph generated by the total ordering of ℤ, then all vertices have indegree and outdegree one (by definition of a total ordering), so there is such an infinite path. Are there some furthe
307.
Physics, optimization, and impossibility(guille.site)
1 points by LolWolf 8y ago|0 comments
308.
Computational bounds for physical design(arxiv.org)
1 points by LolWolf 8y ago|0 comments
309.
by LolWolf 8y ago
Yes, please. The current title is quite misleading: I thought it was a “methods and history of GR tests” survey article when I clicked on it.
310.
by LolWolf 8y ago
> which function is non-differentiable at points? Sorry, this was referring to the construction provided in the paper referenced. I do agree that the title is somewhat misleading, since, when I first read it (and thought, "this is p
311.
by LolWolf 8y ago
> that reference you can't find right now seems rather pertinent? Here it is: https://arxiv.org/pdf/1707.08706.pdf (This isn't quite the one I was thinking of, so I'll dig a little deeper, but it cove
312.
by LolWolf 8y ago
Sure: you can define convexity for a function that is equivalent to the 2nd derivative definition in the case that the function is twice-differentiable, using only the definition of the convexity of a set. Define the epigraph of a functio
313.
by LolWolf 8y ago
Re: "I’m not even sure how one would begin to approximate a highly algorithmic process (e.g., a hash function) using a neural network" You build the circuit corresponding to the function and map it to an NN. Can this be discovered
314.
by LolWolf 8y ago
1. > It's theoretically impossible to guarantee a convergence to global optima using gradient descent if the function is non-convex. This is false. See, e.g., [0][1]. 2. I'm not really sure what the question is here. 3. If your
315.
Glyphosate perturbs the gut microbiota of honey bees(pnas.org)
128 points by LolWolf 8y ago|40 comments
316.
by LolWolf 8y ago
Interesting cycle you've constructed, there.
317.
by LolWolf 8y ago
Eh, I don’t think that many are required. Convergence to a Gaussian is pretty fast (you should check out page 299 of [0]), at four or five a Gaussian is already a quite good approximations. ——— [0] https://www.dartmouth.edu/
318.
by LolWolf 8y ago
Also, gaussians are great approximations for large n, too, since the convolution of any distribution with itself n times (for n "large enough") is close to gaussian (by the CLT. More generally, there are very nice error estimates
319.
by LolWolf 8y ago
"Necessary but not sufficient."
320.
by LolWolf 8y ago
Author here! Feel free to ask any questions :)
321.
Markov processes and the second law(guille.site)
2 points by LolWolf 8y ago|1 comments
322.
by LolWolf 8y ago
Yeah, unsure about what the parent may have intended. I use it as a data IO for stuff that I want to manipulate in Python somewhat-interactively, by having an Excel sheet open on the side as I work in Python.
323.
by LolWolf 8y ago
You can do this with xlwings![0] Would highly recommend. —— [0] https://www.xlwings.org
324.
by LolWolf 8y ago
Not sure what you mean, but my blog is currently generated by Pelican, and on dev server mode, live reload is unnoticeably fast! (To be fair, my blog is also relatively tiny, so it may be a consequence of this). On the other hand, would lov
325.
by LolWolf 8y ago
Author here! Would love to hear any thoughts or ideas you may have (for either this post, or future posts)!
326.
Machine learning, information theory, and tail bounds(guille.site)
3 points by LolWolf 8y ago|1 comments
327.
by LolWolf 8y ago
> Also, most engineering PhD's are bogus because most engineering "research" is actually not deep research. Hello! Engineering researcher here. I toy with things that are both wildly theoretical (information bounds for alg
328.
by LolWolf 8y ago
Hehe, will do!
329.
by LolWolf 8y ago
> For example, can a program decide, given a sorting function implemented in language X, and an algorithm Y, whether or not the sorting function is an implementation of algorithm Y? Assuming the usual things (e.g. that an algorithm Y can
330.
by LolWolf 8y ago
This is a fantastic post! Thank you so much for sharing it. Perhaps another post (which maybe you should write! Or I may (though I'm certainly nowhere as good), or why not both!) might the unifying principle of finding solutions to fix

More ›