Add correct priors on camera properties (mu, sigma, gain)

blinx/hyper_parameters.py

@@ -92,6 +92,8 @@ def __init__(
         g_scale=None,
         mu_loc=None,
         mu_scale=None,
+        sigma_loc=None,
+        sigma_scale=None
     ):
         self.min_y = min_y
         self.num_guesses = num_guesses
@@ -112,6 +114,8 @@ def __init__(
         self.g_scale = g_scale
         self.mu_loc = mu_loc
         self.mu_scale = mu_scale
+        self.sigma_loc = sigma_loc
+        self.sigma_scale = sigma_scale
 
         if sum([r_e_loc is None, r_e_scale is None]) == 1:
             raise RuntimeError("Both r_e_loc and r_e_scale need to be provided")
@@ -121,3 +125,5 @@ def __init__(
             raise RuntimeError("Both g_loc and g_scale need to be provided")
         if sum([mu_loc is None, mu_scale is None]) == 1:
             raise RuntimeError("Both mu_loc and mu_scale need to be provided")
+        if sum([sigma_loc is None, sigma_scale is None]) == 1:
+            raise RuntimeError("Both sigma_loc and sigma_scale need to be provided")

blinx/trace_model.py

@@ -33,11 +33,16 @@ def log_p_parameters(parameters, hyper_parameters):
         )
     if hyper_parameters.g_loc is not None:
         log_p += jnp.log(
-            norm.pdf(parameters.g, hyper_parameters.g_loc, hyper_parameters.g_scale)
+            norm.pdf(parameters.gain, hyper_parameters.g_loc, hyper_parameters.g_scale)
         )
     if hyper_parameters.mu_loc is not None:
         log_p += jnp.log(
-            norm.pdf(parameters.mu, hyper_parameters.mu_loc, hyper_parameters.mu_scale)
+            norm.pdf(parameters.mu_ro, hyper_parameters.mu_loc, hyper_parameters.mu_scale)
+        )
+
+    if hyper_parameters.sigma_loc is not None:
+        log_p += jnp.log(
+            norm.pdf(parameters.sigma_ro, hyper_parameters.sigma_loc, hyper_parameters.sigma_scale)
         )
 
     # sigma is a uniform prior distribution and will add a constant to all models --> we leave it out