Inufu

16 ms·

31.
by Inufu 7y ago
Nice visualizations and explanation! You might want to make it clearer that the agents don't actually receive any visual observations, but rather directly the xy positions of all other agents and objects. This also seems very similar t
32.
by Inufu 7y ago
L7 at Google here (joined as L3 right out of university). Having feedback from L+1 or +2 peers is very important, otherwise it can be very hard for the committee to judge performance. Actually "leading" a project less so - often a
33.
Show HN: AlphaZero Science paper(deepmind.com)
200 points by Inufu 8y ago|89 comments
34.
by Inufu 9y ago
From the article this sounds exactly like my experience playing code golf - coming up with a cool solution, finding out your friend did it in half as many characters, completely reinventing your approach, repeat.
35.
by Inufu 9y ago
We are also doing an AMA on reddit tomorrow: https://www.reddit.com/r/MachineLearning/comments/76xjb5/ama...
36.
by Inufu 9y ago
https://deepmind.com/blog/alphago-zero-learning-scratch/
37.
[Nature] – AlphaGo Zero: Mastering the game of Go without human knowledge(nature.com)
2 points by Inufu 9y ago|2 comments
38.
by Inufu 9y ago
Did you read the article? It's whole argument is that housing in Japan is much cheaper than other cities.
39.
by Inufu 9y ago
Sure, but if you want to actually check for some error condition (say distinguish between file opened ok, file not found, or some other error) - which is the whole point of having an error type to begin with, otherwise you could just use op
40.
by Inufu 9y ago
My main gripe with Rust so far has been the unnecessary profusion of Result<> types, making it hard to process and forward errors. Case in point: the example in the article from the rust documentation that converts errors to strings j
41.
by Inufu 10y ago
It looks like what you are actually timing is the time it takes to start a process - time for either communication method should be on the order of microseconds or less. For better results, start the two processes first, then measure how lo
42.
by Inufu 10y ago
Get a broker account in the country you are living in, then buy Vanguard index funds, e.g. VOO (or local version thereof).
43.
by Inufu 10y ago
tl;dr: Don't use Either<L, R> for error handling, it's too general. Use StatusOr<T>. More details: http://www.furidamu.org/blog/2017/01/28/error-handling-with-...
44.
by Inufu 10y ago
In my experience (Austria and UK), lunch break is part of the 8h, i.e. I get to work at 10:00 and leave at 18:00.
45.
by Inufu 10y ago
There's a good chance that anything on RT is either false or presented in a very misleading way - they are a propaganda outfit for the Kremlin after all. See eg. http://www.economist.com/blogs/graphicdetail/20
46.
by Inufu 10y ago
Presumably the probability density function, https://en.wikipedia.org/wiki/Probability_density_function
47.
by Inufu 10y ago
The expense ratio for VFINX is 0.16%, compared to average yearly returns of 6%. Over 30 years, that's the difference between 5.74x and 5.48x your starting capital. Noticeable, but insignificant compared to the taxes you'll pay.
48.
by Inufu 11y ago
Both YouTube and Netflix usually have local caches, so it's not surprising that popular content plays fine.
49.
by Inufu 11y ago
C++ and Lua.
50.
by Inufu 11y ago
I'd love to put it on a public go server and will try to convince people :) However, this will have to wait until after the match in March, that's our number 1 priority at the moment. There are graphs in the paper showing how it s
51.
by Inufu 11y ago
There will be a live broadcast on our YouTube channel, I think the date will be announced soon.
52.
by Inufu 11y ago
The games were all played on separate days. As Fan Hui mentioned in the video, he changed his strategy after the first game to fight more, so that may explain why it seems his performance changed.
53.
by Inufu 11y ago
The version of AlphaGo used in the paper beats CrazyStone about 99% of the time. Even if we give CrazyStone 4 stones handicap, AlphaGo still beats it 80% of the time.
54.
by Inufu 11y ago
Our paper: http://www.nature.com/nature/journal/v529/n7587/full/nature1... Video from Nature: https://www.youtube.com/watch?v=g-dKXOlsf98&feature=youtu.be Video from us at DeepM
55.
by Inufu 11y ago
More context: http://www.bbc.co.uk/news/technology-35420579 Video from Nature: https://www.youtube.com/watch?v=g-dKXOlsf98&feature=youtu.be Video from us at DeepMind: https://www.youtube
56.
by Inufu 11y ago
I'll just say: Mastering the game of Go with deep neural networks and tree search - http://www.nature.com/nature/journal/v529/n7587/full/nature1...
57.
Mastering the game of Go with deep neural networks and tree search(nature.com)
58 points by Inufu 11y ago|6 comments
58.
by Inufu 11y ago
https://vid.me/65Qw
59.
by Inufu 12y ago
I think you missed the point. Of course you'd never implement add_one that way, but at some point you'll have a function where you want to pass a unique_ptr as an argument, and then your code will happily continue to compile even
60.
by Inufu 12y ago
If you are outside of the USA: https://vid.me/search?q=last%20week%20tonight%20government

More ›