A fast implementation of probabilities, using log-probabilities
A fast implementation of probabilities / positive floating point arithmetic using logarithms.
This library is header only, which means you can copy it and use it with a simple include:
#include "probability/probability.hpp"
We provide aliases for the most common uses:
| Type | Definition |
|---|---|
log_float_t |
Logarithm floating point with value type float |
log_double_t |
Logarithm floating point with value type double |
log_long_double_t |
Logarithm floating point with value type long double |
probability_float_t |
Probability with base type float |
probability_double_t |
Probability with base type double |
probability_long_double_t |
Probability with base type long double |
probability_t |
Alias to probability_double_t |
The library defines a class LogFloatingPoint with 3 template parameters:
T, the value type, used for internal storageulp, the units in the last place, used to define the precision of comparisons.C, the checker type, used to inject methods that verify consistency; the library provides two standard checkers: EmptyChecker (for the log_*_t types above) and ProbabilityChecker (for the probability_*_t types above).By default, all aliases above have ulp = 0 (meaning that the precision equals the machine epsilon of the value type).