tcpekin

8 ms·

31.
by tcpekin 5y ago
I see, I wasn't thinking about kernels much smaller than the image itself. So, if you don't use blocked FFT, it only makes sense to use FFT convolution when your kernel size is > log n?
32.
by tcpekin 5y ago
I have been trying the beta and it's pretty decent. Still alternate between signal, telegram and whatsapp for different groups/people, but the beta definitely seems better that what was before. https://faq.whatsapp.com&
33.
by tcpekin 5y ago
I think often the FFT is used to actually speed up convolution, it isn't slower than computing convolution of your kernel shifted to every pixel position. https://en.wikipedia.org/wiki/Convolution#Fast_convolution_
34.
by tcpekin 5y ago
Fig. 9 of this paper is really interesting if you zoom in. It looks like if the model was not trained on the appropriate class label, it just goes completely off the rails. As previous commenters have noted, I would be very hesitant to use
35.
by tcpekin 5y ago
This happens in numpy as well, and was confusing the first time I saw it: b = np.ones(1) c = np.ones(1) c = c/(2**73) b += c >>> ufunc 'add' output (typecode 'O') could not be coerce
36.
by tcpekin 5y ago
The real problem with these masks is that unless you are quite good at equalizing with your jaw, it will be hard to relieve pressure if you dive down more than 6-8 feet or so.
37.
by tcpekin 5y ago
I think it's just very poorly worded. Sailboats can go faster than the true wind speed, like in the America's Cup, and typically do it do it going upwind. The addition of tacking is not wrong, it is zigzagging upwind, but just ver
38.
by tcpekin 5y ago
$36 a year for 1Password is perhaps the best value/$ I get out of any software I pay for.
39.
by tcpekin 5y ago
I'm not sure that is Germany as a whole though, Berlin yes, but in Munich there are not nearly as many spatis, and from the people that I've interacted with, in smaller towns the only place to get a beer in the evening/night
40.
by tcpekin 5y ago
Does anyone know a way to disable tab scrolling? The main thing that prevents me from switching is the fact that I can't see all my tabs at the same time. They don't continually shrink so I can't use any spatial awareness of
41.
by tcpekin 5y ago
My favorite acid mixes have been Dr. Rubenstein's Quarantine series: https://www.youtube.com/watch?v=0XYcwN4wXhI
42.
by tcpekin 5y ago
I can confirm acid techno is alive and well in Berlin. Lots of good music coming out and DJs playing it.
43.
Index Sets and Parallelism-Preserving Autodiff for Pointful Array Programming(arxiv.org)
1 points by tcpekin 6y ago|0 comments
44.
by tcpekin 6y ago
https://threadreaderapp.com/thread/1374438210315513864.html
45.
by tcpekin 6y ago
I recently read a very good book that was not so much a broad overview, but rather a closer look at the American ambassador and his family in Germany in the 1930s. I can wholeheartedly strongly recommend it. https://en.wikipedia.
46.
by tcpekin 6y ago
I did this with a pair of Focal Alpha 65s (albeit ~2.5 as expensive as the JBLs, and larger too), but I am not sure I can ever go back to smaller computer speakers again. The sound is so so so much better.
47.
by tcpekin 6y ago
What a cool problem to work on - would have never have thought of it. What is the approx. hardware cost?
48.
by tcpekin 6y ago
Could you explain how this works, or point me in the right direction on what to read? I'm interested in not using Transferwise as much if this is better.
49.
by tcpekin 6y ago
I can also recommend looking into voroni clusters for amorphous glasses, as a level 3 type analysis. These can be combined into higher order "medium range order" clusters, which might be what call level 4, but these are all very i
50.
by tcpekin 6y ago
Currently watching it using Streamlink [0] to watch in VLC. This is so exciting! Wishing them, and the rover, all the best in the landing! [0] https://streamlink.github.io/
51.
by tcpekin 6y ago
> What is the common denominator between all metals? I think most people who study this would say the common denominator is that metals lack a band gap and have a very mobile sea of electrons, and therefore good conductivity. Metals can
52.
by tcpekin 6y ago
Happened to me in the exact same way with eBay. User (relatively inactive) for 13+ years, go to sell something (boots lol) with original pics, full description, everything set up correctly, and banned within 5 minutes. No way to appeal, and
53.
by tcpekin 6y ago
In a very similar situation, I posted a pair of "Waxed Flesh" boots on eBay from my 13 year (albeit rarely used) account. They had plenty of pictures, waxed flesh is a very well known term with regards to boot leathers, but within
54.
by tcpekin 6y ago
Someone once posted a website they made in the HN comments that filtered Craigslist/Ebay for the exact old style laptops to reuse... does anyone remember what that was, or what it was called? Was a very nice way to keep on top of the o
55.
by tcpekin 6y ago
We use them for computational imaging reconstruction in electron microscopy.
56.
by tcpekin 6y ago
I happened to use Google Takeout to download everything I had last weekend (~200 Gb)... at first I thought my account was banned and that was the only solace I had, but this definitely is killing my productivity.
57.
by tcpekin 6y ago
Unfortunately for me, I can't access simple analysis since the site can't figure out what country I'm in (no permissions were even asked for), and when I click on the intermediate analysis, the link doesn't seem to be va
58.
by tcpekin 6y ago
I had the exact same experience there about 20 years ago as well with my family... was fantastic.
59.
by tcpekin 6y ago
Here is the actual article, with much more analysis: https://science.sciencemag.org/content/369/6504/689
60.
by tcpekin 7y ago
As mentioned in the article, Nat Comm has been doing this for a few years now. I have found it immensely interesting and helpful for my own research to have access to the peer review files (usually in the Supplementary Information [1]). It

More ›