Changed - the `entropy` function takes frequency counts (multiplicities) as input via the `zk` optional argument - renamed `pk` to `nk` - the signature of the *entropy* function is:
Added - full Bayesian error estimate (from direct computation of the posterior variance of the entropy)
1.7
Changed - rename version.json to package.json - check_pk: elements of input counts array are converted to int32 with no check - entropy() function: NSB estimator needs cardinality `k` as input - entropy() function: non-flat arrays are valid input counts Fixed - robust integration range
1.6.4
Fixed - fix coinformation function for combinations of features - fix from_data for scalar inputs
1.6.1
Changed For the NSB integration: - Find the saddle point maximizing p(alpha | data) - Set the integration range at 4 standard deviations around the saddle point - If the standard deviation is too small, return the entropy posterior mean at the saddle point