NSketch |
The NSketch library aims to provide clean and well documented .Net 2.0 implementation of the most useful sketch-based algorithms. The library is written in C# and targets both Microsoft .Net Framework and Mono.
Sourceforge: NSketch project page
A sketch is a compact yet approximate representation of some data. Most well-studied sketches quantify the level of approximation through the Probably Approximately Correct (PAC) framework . Intuitively, a typical PAC property indicates that the (approximate) output is within a ε error margin of epsilon with a probability greater than P. If exactness is not a requirement, the approximation tradeoff can provide a huge performance gain against amount of limited error.
NSketch v0.1 is available for download here. The version 0.1 includes histograms (naive, sechap, exponential), frequent item selection (lossy counting), bloom filters, fast generic hash function. The library is fully documented (look at the online documentation).
Any feedback on the NSketch library will be greatly appreciated.
Author: Joannès Vermorel