1. kBinCoeff is subject to overfitting. LinearSampling must return -1 whenever calculation hit its boundaries.
2. Investigate whether we can use GMP http://gmplib.org/
I don't think we actually have to use GPM library and long arithmetic.
Current approach: $(sum_i f(i)) / C_L^\ell$
Robust approach: $sum_i (f(i) / C_L^\ell)$.
It should be possible to calc each sumand f(i) / C_L^\ell because f(i) is usually a binomial coef, e.g. you easily get its e-logarithm.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I don't think we actually have to use GPM library and long arithmetic.
Current approach: $(sum_i f(i)) / C_L^\ell$
Robust approach: $sum_i (f(i) / C_L^\ell)$.
It should be possible to calc each sumand f(i) / C_L^\ell because f(i) is usually a binomial coef, e.g. you easily get its e-logarithm.