Merge pull request #182 from jrleja/phot_outlier

PhotoCal model for correlated photometric noise

prospect/fitting/fitting.py

@@ -99,7 +99,8 @@ def lnprobfn(theta, model=None, obs=None, sps=None, noise=(None, None),
         sigma_spec = spec_noise.construct_covariance(**vectors)
     if phot_noise is not None:
         phot_noise.update(**model.params)
-        vectors.update({'phot_unc': obs.get('maggies_unc', None)})
+        vectors.update({'phot_unc': obs.get('maggies_unc', None),
+                        'phot': obs.get('maggies', None)})
 
     # --- Generate mean model ---
     try:

prospect/likelihood/kernels.py

@@ -1,6 +1,6 @@
 import numpy as np
 
-__all__ = ["Kernel", "Uncorrelated", "ExpSquared", "Matern"]
+__all__ = ["Kernel", "Uncorrelated", "ExpSquared", "Matern", "PhotoCal"]
 
 
 class Kernel(object):
@@ -94,8 +94,16 @@ def construct_kernel(self, metric):
         return Sigma
 
 
-class Outliers(Kernel):
-    kernel_params = ['amplitude', 'location']
+class PhotoCal(Kernel):
 
+    ndim = 2
+    npars = 2
+    kernel_params = ['amplitude', 'filter_names']
+
     def construct_kernel(self, metric):
-        raise(NotImplementedError)
+        """ This adds correlated noise in specified bands of photometry
+        """
+        k = np.array([f in self.params["filter_names"] for f in metric])
+        K = k[:, None] * k[None, :]     # select off-diagonal elements
+        return K * self.params["amplitude"]**2
+