Merge branch '20170904-fsaad-log-constants-store'

Author: FA Saad

src/primitives/normal.py

@@ -15,13 +15,20 @@
 # limitations under the License.
 
 from math import lgamma
+from math import log
+from math import pi
 
 import numpy as np
 
 from cgpm.primitives.distribution import DistributionGpm
 from cgpm.utils import general as gu
 
 
+LOG2 = log(2)
+LOGPI = log(pi)
+LOG2PI = LOG2 + LOGPI
+
+
 class Normal(DistributionGpm):
     """Normal distribution with normal prior on mean and gamma prior on
     precision. Collapsed.
@@ -157,39 +164,40 @@ def is_numeric():
 
     @staticmethod
     def calc_predictive_logp(x, N, sum_x, sum_x_sq, m, r, s, nu):
-        mn, rn, sn, nun = Normal.posterior_hypers(
+        _mn, rn, sn, nun = Normal.posterior_hypers(
             N, sum_x, sum_x_sq, m, r, s, nu)
-        mm, rm, sm, num = Normal.posterior_hypers(
+        _mm, rm, sm, num = Normal.posterior_hypers(
             N+1, sum_x+x, sum_x_sq+x*x, m, r, s, nu)
         ZN = Normal.calc_log_Z(rn, sn, nun)
         ZM = Normal.calc_log_Z(rm, sm, num)
-        return -.5 * np.log(2*np.pi) + ZM - ZN
+        return -.5 * LOG2PI + ZM - ZN
 
     @staticmethod
     def calc_logpdf_marginal(N, sum_x, sum_x_sq, m, r, s, nu):
-        mn, rn, sn, nun = Normal.posterior_hypers(
+        _mn, rn, sn, nun = Normal.posterior_hypers(
             N, sum_x, sum_x_sq, m, r, s, nu)
         Z0 = Normal.calc_log_Z(r, s, nu)
         ZN = Normal.calc_log_Z(rn, sn, nun)
-        return -(N/2.) * np.log(2*np.pi) + ZN - Z0
+        return -(N/2.) * LOG2PI + ZN - Z0
 
     @staticmethod
     def posterior_hypers(N, sum_x, sum_x_sq, m, r, s, nu):
         rn = r + float(N)
         nun = nu + float(N)
         mn = (r*m + sum_x)/rn
         sn = s + sum_x_sq + r*m*m - rn*mn*mn
-        if sn == 0: sn = s
+        if sn == 0:
+            sn = s
         return mn, rn, sn, nun
 
     @staticmethod
     def calc_log_Z(r, s, nu):
         return (
-            ((nu + 1.) / 2.) * np.log(2)
-            + .5 * np.log(np.pi)
-            - .5 * np.log(r)
-            - (nu/2.) * np.log(s)
-            + lgamma(nu/2.0))
+            ((nu + 1.) / 2.) * LOG2
+            + .5 * LOGPI
+            - .5 * log(r)
+            - (nu/2.) * log(s)
+            + lgamma(nu/2.))
 
     @staticmethod
     def sample_parameters(m, r, s, nu, rng):