avaku

14 ms·

121.
by avaku 13y ago
Didn't see any problem with gevent example... Looks like the standard parallel code.
122.
by avaku 13y ago
Can I come up with a bit of criticism? This book does provide a great description of the details of the algorithm inner workings (very cute demons too). However, after reading this chapter (sorry I haven't looked at the other ones), th
123.
by avaku 13y ago
It's funny because the video is shot in China...
124.
by avaku 13y ago
I agree too. I don't regret a single second. True, the business didn't work out. But what I've learned is so much more, than if I would keep working a perm job... The startup experience changed my life for the better. It will
125.
by avaku 13y ago
First hand exp! :)
126.
by avaku 13y ago
Thanks, I'll try to write something up...
127.
by avaku 13y ago
I agree, almost nobody writes: "I've been warned, but I pursued my dream anyway, and then I failed miserably." I did.
128.
by avaku 13y ago
There were lots of woman scientists in USSR, and I don't think there was gender-related discrimination... Much fewer than men, but probably not for the reasons that are now popular to attribute to this phenomenon. Just saying...
129.
by avaku 13y ago
This can be stacked in multiple levels, cool :) JVM on Python on JVM on Python on JVM...
130.
by avaku 13y ago
Some other people do!
131.
by avaku 13y ago
I just implemented math from a textbook. Obviously, you will need to break down your song into smaller chunks. Then compare the frequencies of these chunks. When sufficiently many chunks are "close" (you'll need to define som
132.
by avaku 13y ago
Be careful not to get sued like that guy who posted an article about detecting songs and got sued by Shazam, even though everybody knows about FFT (I've been doing something like that for my BSc)
133.
by avaku 13y ago
Was the project shut down? (the website is not working, sorry if I haven't missed that in the article)
134.
by avaku 13y ago
It was just a strange phrase. The idea with hitting all exchanges at once is not bad at all...
135.
by avaku 13y ago
Katsuyama says: "When I'm trying to place a large order, the price goes up!" So, what's wrong with that? Have we cancelled the laws of supply and demand now? Let's beat up Adam Smith now!
136.
by avaku 13y ago
KNN is not nearly as good as deeper probabilistic models. For example, if movie description mentions different subjects like "comedy" and "family", the model would not be able to differentiate them. It would only have a
137.
by avaku 13y ago
This is a trick. The correct thing to say to them - we don't have passwords because they are hashed and salted. Then you successfully pass the security audit :)
138.
by avaku 13y ago
He doesn't even sound..
139.
by avaku 13y ago
It's easy for us to judge now. But if you read this article when it came out, you wouldn't be so sure. This is the reason why not many analysts publish reports that are contrary to "common" opinion. Because if everyone i
140.
by avaku 13y ago
Thanks HN! Although, there is an element of overcompensation in finance, as well as a lot of overconfident people with more testosterone than knowledge, it still does provide hell of a lot of services. I'm happy to see that the HN crow
141.
by avaku 13y ago
I love it!
142.
by avaku 13y ago
You could set up a cost function to account for closeness to being hit (a.k.a. decision theory approach)
143.
by avaku 13y ago
A similar app was recently in the top chart in "productivity" section. It's called feeltracker, and it has historical graphs: https://www.feeltracker.com
144.
by avaku 13y ago
Obvious: comments, twits... Any non-obvious ideas?
145.
Show HN: To what data I can apply my news search engine, other than news?(readrz.com)
1 points by avaku 13y ago|1 comments
146.
by avaku 13y ago
Genius :)
147.
by avaku 13y ago
I knew it!
148.
by avaku 13y ago
I would like to say that the Russian apartment bombings, although not officially proven to originate from the government, upon thorough investigation appear to have a very significant probability of being an actual brutal false flag. In add
149.
by avaku 13y ago
Great explanation. I would like to add to this, that held-out data is often used in Bayesian learning too - for example, in cases when you intentionally over-specify the model (adding more parameters than might be needed) because you don&#x
150.
by avaku 13y ago
Eventually, not everyone will want to pay "cash" (irreversible transfer) for everything. Additional security and services layer will be added to Bitcoin ecosystem, and the fee for making indirect transfers through these intermedia

More ›