Bayesian Filter Library / News: Recent posts

libbpfl is a general purpose C++ library for Bayesian Filtering. Essentially Bayesian Filtering is a way of having a program learn to categorize information from a specific user through pattern recognition. This release has some fairly big changes, mainly to the interface, some classes were renamed to be a little more decriptive. Internally the calculations are done with arbitrary precision ratios, for more exact score calculation.

I've begun making and incorporating design plans for the next minor level release (0.2.0). So far, I've incorparated the CLN Numeric library for exact computation, this was really required since if large vectors of scores are multiplies together they can lose precision to the point of becoming zero, which totally kills the whole scoring process. Using CLN's Ratio types, provides exact computations which are only converted to float's on output.... read more

This came pretty quickly after the last, but in the development of bmpflat, I realized I had to enable the main classes to have virtual bases. The reason why I released this patch is because I want to have a library that will be fully compatible with bmpflat when it's ready.

This is just a very minor interface change. See the DhangeLog for details.

This is the first release of libbpfl, it is ALPHA, I'm pretty confident that it works well, but the interface isn't refined, and the documentation is quite lacking. Please feel free to contact me with any questions or problems.