You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I've been extending Utilities/Probability for a little bit with additional probability distribution sampler functions and have a couple of gripes:
As far as I can tell Utilities/Probability is currently a normal distribution single-sampler and doesn't actually deal with larger concepts involving probability. This makes the module name appear to be a misnomer. I propose that it be renamed Utilities/Distributions, Utilities/Stats, or something similar.
Further, the inclusion of the Distribution suffix in the module's functions is redundant. We already know that the module is about "Probability" and so exported function names should be obvious.
BoxMuller is an implementation detail that should not be exposed to module consumers.
Using single-sampler functions, functions that only sample one value from the distribution in question, does not allow us to abuse the strengths of accept-reject sampling methods to their fullest.
The text was updated successfully, but these errors were encountered:
JCRPaquin
changed the title
Issues with Utilities/Probability
Issues with Utilities/Probability
Mar 22, 2017
I've been extending
Utilities/Probability
for a little bit with additional probability distribution sampler functions and have a couple of gripes:As far as I can tell
Utilities/Probability
is currently a normal distribution single-sampler and doesn't actually deal with larger concepts involving probability. This makes the module name appear to be a misnomer. I propose that it be renamedUtilities/Distributions
,Utilities/Stats
, or something similar.Further, the inclusion of the
Distribution
suffix in the module's functions is redundant. We already know that the module is about "Probability" and so exported function names should be obvious.BoxMuller
is an implementation detail that should not be exposed to module consumers.Using single-sampler functions, functions that only sample one value from the distribution in question, does not allow us to abuse the strengths of accept-reject sampling methods to their fullest.
The text was updated successfully, but these errors were encountered: