Bandits are probably one of the most underrated machine learning algorithms. One possible application is recommendation systems. Shameless self promotion. I wrote an article about it: https://towardsdatascience.com/how-not-to-sort-by-popularity...

We are just exporting the matrices. Nothing fancy.

I did something similar in Javascript. My implementation uses the Runge-Kutta algorithm, that is more stable. My GUI does not allow planet creation, but it shows several stable configurations, including the 13 recently found three body planar orbit configurations.

http://jbochi.github.io/planets/ http://suki.ipb.ac.rs/3body/

Riak CS is S3 compatible.

I have a very similar problem and end up writing a python script that copies the files to cassandra and Nginx + Lua to serve them using lua-resty-cassandra. The read/write throughput is still not as good as I was expecting though.

This article is the best introduction to the topic in my opinion: http://steve.vinoski.net/pdf/IC-Rediscovering-Distributed-Sy...

Another good reference for understanding distributed databases is: http://cs-www.cs.yale.edu/homes/dna/papers/abadi-pacelc.pdf

It would be interesting to see the same data broken down by year. Lua, Clojure, and Go are at about the same position, but I wonder which one is gaining traction. Probably Go.

This blog post is a nice survey about other interesting probabilistic data structures: http://highlyscalable.wordpress.com/2012/05/01/probabilistic...

It's hard to believe that JavaScript and Transact-SQL have more or less the same popularity (1.569% against 1.559%). That's why I don't care about this index anymore.

Barnes-Hut is O(n log n) http://en.wikipedia.org/wiki/Barnes%E2%80%93Hut_simulation

Yes, once you start cutting off long range effects you can improve on the strict O(N^2) limitations.

It's not cutting off. It's replacing many long-range effects with a single one weighted more (and done it carefully so that the floating-point error is literally smaller than you could tell).

There's even better methods that take O(n) time, like the fast multipole expansion ones.

Btw this goes into numerical analysis, and O(-) is important but so is error convergence. I imagine these big-O figures are for a constant error, but I figure the asymptotic may change behavior as the other parameter changes.

One could also use FFTs and the particle-mesh method or a TreePM method. Basically, the particle masses are deposited on a fine mesh. Poisson's equation gets solved using FFTs and then the particle accelerations are computed by interpolating on the gradient of the potential field.

There is a nice big list here: http://cstheory.stackexchange.com/questions/1168/what-papers...