Speed up core.prior classes
Extended from #411 (closed)
-
-
Profile all of the priors which use
scipy
: LogNormal, Exponential, StudentT, Logistic, Cauchy, Gamma.
-
Profile all of the priors which use
-
-
See if we can make them run faster by using analytical solutions in the
rescale
,prob
,logprob
,cdf
methods.- Uniform: merged and dealt with in #411 (closed). O(1)-O(2) faster.
- LogNormal: 8x to 20x faster.
- Exponential: 5x to 50x faster.
- StudentT: 3x to 40x faster.
- Logistic: 5x to 60x faster.
- Cauchy: 2x to 500x faster.
- Gamma: 2x to 600x faster.
- Beta: 2x to 300x faster.
-
See if we can make them run faster by using analytical solutions in the
-
- Add unit tests to ensure that calculations are correct.
(Biggest speed-up is when inputting (float
, int
), smaller speed-up is inputting np.ndarray
)