Changeset 27766 for issm/trunk-jpl/src/m/classes/m1qn3inversion.py

Timestamp:

05/19/23 08:08:54 (22 months ago)

Author:

Mathieu Morlighem

Message:

CHG: adding new field to m1qn3 inversions (initial cost function reduction)

File:

• issm/trunk-jpl/src/m/classes/m1qn3inversion.py (modified) (5 diffs)

issm/trunk-jpl/src/m/classes/m1qn3inversion.py

@@ -19 +19 @@
         if not len(args):
             print('empty init')
-            self.iscontrol = 0
-            self.incomplete_adjoint = 0
-            self.control_parameters = np.nan
-            self.control_scaling_factors = np.nan
-            self.maxsteps = 0
-            self.maxiter = 0
-            self.dxmin = 0.
-            self.gttol = 0.
-            self.cost_functions = np.nan
+            self.iscontrol                   = 0
+            self.incomplete_adjoint          = 0
+            self.control_parameters          = np.nan
+            self.control_scaling_factors     = np.nan
+            self.maxsteps                    = 0
+            self.maxiter                     = 0
+            self.dxmin                       = 0.
+            self.dfmin_frac                  = 0.
+            self.gttol                       = 0.
+            self.cost_functions              = np.nan
             self.cost_functions_coefficients = np.nan
-            self.min_parameters = np.nan
-            self.max_parameters = np.nan
-            self.vx_obs = np.nan
-            self.vy_obs = np.nan
-            self.vz_obs = np.nan
-            self.vel_obs = np.nan
-            self.thickness_obs = np.nan
+            self.min_parameters              = np.nan
+            self.max_parameters              = np.nan
+            self.vx_obs                      = np.nan
+            self.vy_obs                      = np.nan
+            self.vz_obs                      = np.nan
+            self.vel_obs                     = np.nan
+            self.thickness_obs               = np.nan
 
             self.setdefaultparameters()
@@ -71 +72 @@
         s += '{}\n'.format(fielddisplay(self, 'maxiter', 'maximum number of Function evaluation (forward run)'))
         s += '{}\n'.format(fielddisplay(self, 'dxmin', 'convergence criterion: two points less than dxmin from eachother (sup - norm) are considered identical'))
+        s += '{}\n'.format(fielddisplay(self, 'dfmin_frac', 'expected reduction of J during the first step (e.g., 0.3=30% reduction in cost function)'))
         s += '{}\n'.format(fielddisplay(self, 'gttol', '||g(X)||/||g(X0)|| (g(X0): gradient at initial guess X0)'))
         s += '{}\n'.format(fielddisplay(self, 'cost_functions', 'indicate the type of response for each optimization step'))
@@ -108 +110 @@
         #m1qn3 parameters
         self.dxmin = 0.1
+        self.dfmin_frac = 1.
         self.gttol = 1e-4
 
@@ -142 +145 @@
         md = checkfield(md, 'fieldname', 'inversion.maxiter', 'numel', [1], '>=', 0)
         md = checkfield(md, 'fieldname', 'inversion.dxmin', 'numel', [1], '>', 0.)
+        md = checkfield(md, 'fieldname', 'inversion.dfmin_frac', 'numel', [1], '>=', 0., '<=', 1.)
         md = checkfield(md, 'fieldname', 'inversion.gttol', 'numel', [1], '>', 0.)
         md = checkfield(md, 'fieldname', 'inversion.cost_functions', 'size', [num_costfunc], 'values', supportedcostfunctions())
@@ -169 +173 @@
         WriteData(fid, prefix, 'object', self, 'class', 'inversion', 'fieldname', 'maxiter', 'format', 'Integer')
         WriteData(fid, prefix, 'object', self, 'class', 'inversion', 'fieldname', 'dxmin', 'format', 'Double')
+        WriteData(fid, prefix, 'object', self, 'class', 'inversion', 'fieldname', 'dfmin_frac', 'format', 'Double')
         WriteData(fid, prefix, 'object', self, 'class', 'inversion', 'fieldname', 'gttol', 'format', 'Double')
         WriteData(fid, prefix, 'object', self, 'class', 'inversion', 'fieldname', 'cost_functions_coefficients', 'format', 'DoubleMat', 'mattype', 1)