Update 'python/bayesian/Bayesian_Zahra.py'

python/bayesian/Bayesian_Zahra.py

@@ -90,11 +90,6 @@ def unpack_phi_mono(phi):
     
     return p, a, b, s, k, th
 
-# Priors 
-# Beta prior on p centered at empirical rate
-TAU = 25.0   # reduce to ~5 if you want it weaker
-alpha = max(TAU * float(p_emp), 1e-6)
-beta  = max(TAU * (1.0 - float(p_emp)), 1e-6)
 
 # Weak log-normal shrinkage on positive parameters
 def nlog_lognormal(x, mu, sigma, eps=1e-12):
@@ -116,27 +111,7 @@ def neg_post_phi_mono_WITH_CONST_REG(phi, X, y):
     # Prior on p ~ Beta(alpha, beta)
     npr_p = -((alpha-1)*np.log(p + eps) + (beta-1)*np.log(1 - p + eps))
 
-    # Regularization (weak priors)
-    # AE median m1: use AE median if present; otherwise overall median.
-    if (y == 1).any():
-        m1 = np.median(X[y == 1])
-    else:
-        m1 = np.median(X)
-
-    reg = 0.0  # total penalty starts at zero
-
-    # 1) Gamma shape k (>0): very weak prior centered at 2 (σ=1.2).
-    reg += nlog_lognormal(k, mu=np.log(2.0), sigma=1.2)
-
-    # 2) Beta-Prime shape b (>0): same weak prior.
-    reg += nlog_lognormal(b, mu=np.log(2.0), sigma=1.2)
-
-    # 3) Beta-Prime scale s (>0): center near AE median (tighter σ=0.5).
-    reg += nlog_lognormal(s, mu=np.log(max(m1, 1e-6)), sigma=0.5)
-
-    # 4) Left-tail gap delta = a - k (>0): center around ~1.5 (σ=0.5)
-    delta = a - k
-    reg += nlog_lognormal(delta, mu=np.log(1.5), sigma=0.5)
+    
 
     # Keep theta away from the boundary: Beta(3,3) on r = th/th_cap
     thcap = theta_max(a, b, k, s)
@@ -146,7 +121,7 @@ def neg_post_phi_mono_WITH_CONST_REG(phi, X, y):
     else:
         npr_r = 0.0
 
-    return nll + npr_p + reg + npr_r
+    return nll  +npr_r
 
 # Initialization 
 def init_phi(X, y):