trott

11 ms·

91.
by trott 3y ago
Agda's been around for 24 years, according to Wikipedia, and it hasn't exactly revolutionized software development: I'm not aware of any notable software written in it, or of anyone so much as winning a programming competitio
92.
by trott 3y ago
Prolog is Turing-complete, and Datalog isn't. In light of this, I think it's a bit of a stretch to call it a (logic) programming language. But people seem to do that.
93.
by trott 3y ago
I have no doubt that visual is better than text-only and austere, by any metric, except one: filtering out those with less commitment. Laundau himself didn't have much patience for struggling to understand dense language, according to
94.
Using scaling laws to predict when a scaled-up GPT-4 becomes superhuman(old.reddit.com)
1 points by trott 3y ago|0 comments
95.
Learning Logic Programs by Explaining Failures(arxiv.org)
1 points by trott 5y ago|0 comments
96.
by trott 5y ago
> is Prolog a good language for doing statistics or ML or scientific computing? Not really, but the interesting thing about Prolog is ILP (inductive logic programming). ILP does things current deep learning can not: learning compact mode
97.
by trott 5y ago
I meant "zero-mean normal distribution".
98.
by trott 5y ago
"Independence" is defined for random variables, so "randomness" is implied. However, this approximate orthogonality property depends on the distribution. It holds for the normal distribution and many others, although not
99.
by trott 5y ago
> His explanation is "it only looks like it's moving, due to the zoom". This paper, as well as the pilot that took that video disagrees. Interview with Chad Underwood, the pilot that took the FLIR Video Mick West's cl
100.
Explained: New Navy UFO Videos(youtube.com)
27 points by trott 5y ago|6 comments
101.
by trott 5y ago
I'm curious about how much of this apparent low dimensionality is explained by (1) the physical proximity of the neurons being recorded, (2) poverty of the stimuli (just 4 sequences in this paper, if I'm not mistaken)
102.
by trott 5y ago
Something to keep in mind though is that in a high-dimensional space, approximate orthogonality of independent vectors is almost guaranteed.
103.
“Pyramid UFO” – New Footage. It's Just Bokeh, Not a Pyramid(youtube.com)
52 points by trott 5y ago|4 comments
104.
by trott 5y ago
> Wait... How can 1 Hidden layer be called DEEP neural network? Yes. They describe a general algorithm, but show results for shallow nets. So it's unclear if this works well in general.
105.
Learning large logic programs by going beyond entailment(arxiv.org)
2 points by trott 5y ago|0 comments
106.
by trott 6y ago
This is currently #1 on the front page, but I think HN is celebrating the defeat of GPUs prematurely. The architectures used in this paper have just one hidden activation layer. It remains to be seen whether the ideas in SLIDE are applicabl
107.
Turning 30: New Ideas in Inductive Logic Programming(arxiv.org)
6 points by trott 6y ago|0 comments
108.
by trott 6y ago
OP has the freedom to choose his job and industry.
109.
by trott 6y ago
> The software I build is useless. Isn't this up to you?
110.
by trott 6y ago
> During the first surge (ie, before June, 2020), 533 381 people were tested, of whom 11 727 (2.20%) were PCR positive The conclusion relies on the accuracy of the tests. But what if these tests had a 0.5% false-positive rate? This would
111.
by trott 6y ago
> It included only two other people, both young graduate students in his lab at the university. It made no products. It had no plans to make a product. And its website offered nothing but a name, DNN-research, which was even less invitin
112.
by trott 6y ago
> Would Amazon be possible without trucks? How about Walmart? Not Walmart, but Amazon makes even more sense in a horse-driven economy: First, you don't have to ride to the store and back -- mailing a small item is much cheaper than
113.
by trott 6y ago
> I just wonder why Ada didn't get the credit it deserved as being absolutely bulletproof It's not. All languages make trade-offs in the performance-convenience-safety-etc space, and Ada's choice is not "100% safety&q
114.
by trott 6y ago
> I don't completely understand this: NFNet is slower than its competitors on this benchmark, but they claim higher efficiency. Take a look at F1 and B7: They have the same accuracy, but F1 is smaller and much faster.
115.
by trott 6y ago
> The deepmind paper NFNet-F4+ you were talking about also has external training data. @dheera and I did not mention NFNet-F4+. All models, tables, figures and numbers that we did mention resulted from training on ImageNet alone.
116.
by trott 6y ago
For ConvNets, the memory use of the models themselves is pretty modest. For example, even with 0.5B parameters, with FP32, weights+gradients+momentum should use just 6GB (unless your framework sucks, or you have extra overhead from distribu
117.
by trott 6y ago
> No it hasn't https://paperswithcode.com/sota/image-classification-on-imag ... We were talking about models trained on ImageNet, specifically about the trade-off between accuracy and FLOPs. But the higher-accur
118.
by trott 6y ago
> It appears to me at first glance that NFNet has not achieved SOTA at inference. It has, for larger models (F1 vs B7). See Fig 4 in the Appendix.
119.
by trott 6y ago
I saw ads for $149/sample. Probably not available to consumers yet. But the general trend is precipitous decline (10000000x)
120.
by trott 6y ago
It's a good point. I'm just saying that this coming tech revolution is another issue to consider.

More ›