Merge pull request #120 from MiraGeoscience/GEOPY-2518

GEOPY-2518: Crash on no-upper bound for MVI within a joint process

Author: domfournier

simpeg/directives/_vector_models.py

@@ -129,7 +129,7 @@ def endIter(self):
         model = self.invProb.model.copy()
 
         # Project bounds
-        if self.mode == "cartesian" and np.any(~np.isinf(self.opt.upper)):
+        if self.mode == "cartesian":
             vec_model = []
             indices = []
             upper_bound = []
@@ -140,12 +140,12 @@ def endIter(self):
                 indices.append(mapping.indices)
 
             amplitude = np.linalg.norm(np.vstack(vec_model), axis=0)
-            upper_bound = np.hstack(upper_bound)
-            upper_bound = np.max(upper_bound[~np.isinf(upper_bound)])
+            upper_bound = np.linalg.norm(np.vstack(upper_bound), axis=0)
             out_bound = amplitude > upper_bound
 
             if np.any(out_bound):
                 scale = upper_bound / amplitude
+
                 for ind in indices:
                     vec = model[ind]
                     vec[out_bound] *= scale[out_bound]
@@ -165,16 +165,19 @@ def endIter(self):
             vec_ref = []
             indices = []
             mappings = []
+            upper_bound = []
             for reg in self.regularizations.objfcts:
                 mappings.append(reg.mapping)
                 vec_model.append(reg.mapping * model)
                 vec_ref.append(reg.mapping * reg.reference_model)
+                upper_bound.append(reg.mapping * self.opt.upper)
                 mapping = reg.mapping.deriv(np.zeros(reg.mapping.shape[1]))
                 indices.append(mapping.indices)
 
             indices = np.hstack(indices)
             nC = mapping.shape[0]
             vec_model = cartesian2spherical(np.vstack(vec_model).T)
+            upper_bound = np.linalg.norm(np.vstack(upper_bound), axis=0)
             vec_ref = cartesian2spherical(np.vstack(vec_ref).T).flatten()
             model[indices] = vec_model.flatten()
 
@@ -218,7 +221,9 @@ def endIter(self):
             self.opt.lower[indices] = np.kron(
                 np.asarray([0, -np.inf, -np.inf]), np.ones(nC)
             )
-            self.opt.upper[indices[nC:]] = np.inf
+            self.opt.upper[indices] = np.r_[
+                upper_bound, np.ones_like(indices[nC:]) * np.inf
+            ]
 
             updates = {}
             for misfit in self.misfits: