dllu

20 ms·

241.
by dllu 9y ago
Even a page [1] made by Jimmy Wales himself was deleted 22 minutes after creation [2]. It was later undeleted and nominated for deletion again, sparking a rather lengthy debate [3]. [1] https://en.wikipedia.org/wiki/Mzo
242.
by dllu 10y ago
Every Cruise Automation Chevy Bolt is using two Velodyne HDL-32E, each of which has a list price of $30,000. Price aside, Velodyne is really struggling to produce these units. Some recent HN comments have mentioned lead times of 40 weeks [1
243.
by dllu 10y ago
All evidence from prestigious algorithmic contests also indicate this. For example, the ACM ICPC world finals [0] has been dominated by Russia and China since 2000. Russians also rank highly in Topcoder [1]. American competitive programmer
244.
by dllu 10y ago
They also make stabilized boats. [0] Truly fascinating company. [0] https://news.ycombinator.com/item?id=8553204
245.
by dllu 10y ago
Shaking things often fixes problems. This reminds me of the following story: > [O]ne of the earliest [applications] of dither came in World War II. Airplane bombers used mechanical computers to perform navigation and bomb trajectory calc
246.
by dllu 10y ago
Here's a hilarious design for a keyboard with a key for most Japanese symbols: https://japan.googleblog.com/2010/04/google.html It's an April Fool's joke, obviously. For Japanese you can use romaji
247.
by dllu 11y ago
The developers probably use whatever window manager they want. I'm personally a fan of i3 also. But there is nothing wrong with keeping the frontend machine used in this Go match in default Ubuntu desktop environment since its only pur
248.
by dllu 11y ago
Failure to generalize is not always caused by overfitting. Even if there is no overfitting, deep neural networks seem to learn a surprisingly discontinuous function. Consequently, in rare cases, they can misclassify things with great confid
249.
by dllu 11y ago
Yes, this is the separability property of the Gaussian. It makes it easy to compute Gaussian blur as compared to, say, the circular pillbox filter (lens blur or bokeh simulation) because you can just convolve it different times separately i
250.
by dllu 11y ago
One of the remarkable things about the Gaussian function is that its Fourier transform is also a Gaussian. A Gaussian blur means to convolve a function with the Gaussian. Convolution in the frequency (Fourier) domain is multiplication. Sinc
251.
by dllu 11y ago
After reading this I decided to go out and take pictures of moss. Here's a picture I took: http://i.imgur.com/rovioAa.jpg near 49.055969, -122.238218. It's quite relaxing but my tripod got dirty due to the mud.
252.
by dllu 11y ago
Most structure from motion or monocular SLAM algorithms use some kind of feature tracking and then estimate a dense reconstruction after they optimize the locations of all the cameras and all the features. But it's not the only way. Th
253.
by dllu 11y ago
Considering that -272 C is very close to absolute zero, it makes perfect sense. For example, 0.1 K and 0.5 K are both written as -273 C, but the latter is five times as hot as the former. For best consistency, we ought to use logarithmic Ke
254.
by dllu 11y ago
The aim of cmix is to use as many context mixing models as is necessary to get the best possible compression ratio as an academic exercise. Little attempt was made to make it faster or use less memory. That said, the author recently communi
255.
by dllu 11y ago
A few months ago I created my own Markdown-like lightweight markup language supporting LaTeX equations, which can be compiled to either HTML or LaTeX: http://lawrence.lu/programming/dllup/
256.
by dllu 11y ago
According to the Forbes list of billionaires: http://www.forbes.com/billionaires/list/#version:static_sear... There are currently four Microsoft billionaires (probably fewer at the time referred to in the article)
257.
by dllu 11y ago
It's really cool that Poisson disc sampling is used to generate the points. Here, best-candidate sampling is used, based on [1]. But have you considered using the possibly faster Bridson's algorithm [2]? Using Harris corners to an
258.
by dllu 12y ago
At that level of accuracy, however, you will find that the rotation of the Earth itself doesn't keep time that well. The Shortt-Synchronome clock [0] was the most accurate pendulum clock, known to have an error of less than 1 second pe
259.
by dllu 12y ago
Looks like they got an American Millenial to design their quiz: http://i.imgur.com/j4b73bz.png
260.
by dllu 12y ago
I created a similar thing for my personal website: http://www.dllu.net/programming/dllup/ which handles math using svgtex instead of clientside MathJax for faster rendering. It also compiles to both html5 and LaTe
261.
by dllu 12y ago
I wrote my own lightweight markup language, similar to Markdown, that compiles to both HTML5 and LaTeX, supports LaTeX equations (through svgtex, i.e. serverside MathJax), references, captioned figures with semantically correct <figure&g
262.
Show HN: Dllup markup language – Lightweight markup for mathy blogs(dllu.net)
2 points by dllu 12y ago|1 comments
263.
Show HN: I wrote my own lightweight markup language for mathy blogs
9 points by dllu 12y ago|1 comments
264.
by dllu 12y ago
Output-sensitive algorithms are pretty interesting in general. Previous to Chan's algorithm there was the Kirkpatrick-Seidel algorithm ( https://en.wikipedia.org/wiki/Kirkpatrick%E2%80%93Seidel_alg... ), which is mo
265.
by dllu 12y ago
Maglev also beat that speed many years ago: https://en.wikipedia.org/wiki/Land_speed_record_for_rail_veh... In 2003 the Japanese maglev test train running on the same test track reached 581 km/h. It was not until
266.
by dllu 12y ago
There are some numerical instabilities in my code. Theoretically the integrator I'm using (the RK4 [0]) should be a lot more accurate than the Euler's method that OP is using, but I probably have a poor choice of parameters and&#x
267.
by dllu 12y ago
I did this about a year ago: http://www.dllu.net/dp/ It barely exceeds 100 lines of js though... I have a whole bunch of similar physics simulators here: http://www.dllu.net/programming/physics&#x2
268.
by dllu 13y ago
As long as you qualify, results from each round do not make a difference in subsequent rounds. [0] [0] https://code.google.com/codejam/faq.html
269.
by dllu 13y ago
It just takes a bit of practice. Once you've seen a certain type of problem before, it is easy to identify other problems that can be somehow reduced to a known problem. Speed comes with practice too... soon, basic constructs like for-
270.
by dllu 13y ago
There's not really much difference between the top 1000 places, since all that matters is solving all 4 problems. This is the qualification round that goes on for 27 hours, so there is not much pressure to solve the problems fast. The

More ›