select general mat data only inside the domain for training

pinnicle/modeldata/general_mat_data.py

@@ -29,16 +29,37 @@ def load_data(self, domain=None, physics=None):
         # Reading matlab data
         data = load_mat(self.parameters.data_path)
 
+        # pre load x, y, then use domain.bbox to further get the inflag
+        X = {}
+
         # load the coordinates
         for k, v in self.parameters.X_map.items():
             if v in data:
-                self.X_dict[k] = data[v]
+                X[k] = data[v]
             else:
                 print(f"{v} is not found in the data from {self.parameters.data_path}, please specify the mapping in 'X_map'")
 
+        # use the order in physics.input_var to determine x and y names
+        if physics:
+            xkeys = physics.input_var[0:2]
+        else:
+            xkeys = list(X.keys())
+
+        # get the bbox from domain, set the rectangle, works for both static and time dependent domain
+        if domain:
+            bbox = domain.bbox()
+            # set the flag based on the bbox region
+            boxflag = (X[xkeys[0]]>=bbox[0][0]) & (X[xkeys[0]]<=bbox[1][0]) & (X[xkeys[1]]>=bbox[0][1]) & (X[xkeys[1]]<=bbox[1][1])
+        else:
+            boxflag = np.ones_like(X[xkeys[0]], dtype=bool)
+
+        # load the coordinates
+        for k in X.keys():
+            self.X_dict[k] = X[k][boxflag].flatten()[:,None]
+
         # load all variables from parameters.name_map
         for k in self.parameters.name_map:
-            self.data_dict[k] = data[self.parameters.name_map[k]]
+            self.data_dict[k] = data[self.parameters.name_map[k]][boxflag].flatten()[:,None]
 
     def plot(self, data_names=[], vranges={}, axs=None, **kwargs):
         """ TODO: scatter plot of the selected data from data_names

pinnicle/modeldata/h5_data.py

@@ -40,22 +40,20 @@ def load_data(self, domain=None, physics=None):
 
         # use the order in physics.input_var to determine x and y names
         if physics:
-            xkey = physics.input_var[0]
-            ykey = physics.input_var[1]
+            xkeys = physics.input_var[0:2]
         else:
-            xkey = 'x'
-            ykey = 'y'
+            xkeys = list(X.keys()) 
 
-        # get the bbox from domain, set the rectangle 
+        # get the bbox from domain, set the rectangle, works for both static and time dependent domain
         if domain:
             bbox = domain.bbox()
             # set the flag based on the bbox region
-            boxflag = (X[xkey]>=bbox[0][0]) & (X[xkey]<=bbox[1][0]) & (X[ykey]>=bbox[0][1]) & (X[ykey]<=bbox[1][1])
+            boxflag = (X[xkeys[0]]>=bbox[0][0]) & (X[xkeys[0]]<=bbox[1][0]) & (X[xkeys[1]]>=bbox[0][1]) & (X[xkeys[1]]<=bbox[1][1])
         else:
-            boxflag = np.ones_like(X[xkey], dtype=bool)
+            boxflag = np.ones_like(X[xkeys[0]], dtype=bool)
 
         # load the coordinates
-        for k in self.parameters.X_map.keys():
+        for k in X.keys():
             self.X_dict[k] = X[k][boxflag].flatten()[:,None]
 
         # load all variables from parameters.name_map

tests/test_data.py

@@ -140,19 +140,64 @@ def test_MatData():
     icoord = data_loader.get_ice_coordinates()
     assert icoord.shape == (3192, 2)
 
-    hp["X_map"] = {"x":"x"}
+    hp["X_map"] = {"notload":"t", "y":"y"}
     p = SingleDataParameter(hp)
     data_loader = MatData(p)
     data_loader.load_data()
+    assert ("y" in data_loader.X_dict)
+    assert ("t" not in data_loader.X_dict)
+    assert ("notload" not in data_loader.X_dict)
+
+def test_MatData_domain():
+    filename = "flightTracks.mat"
+    expFileName = "fastflow_CF.exp"
+    repoPath = os.path.dirname(__file__) + "/../examples/"
+    appDataPath = os.path.join(repoPath, "dataset")
+    path = os.path.join(appDataPath, filename)
+
+    hp = {}
+    hp["data_path"] = path
+    hp["data_size"] = {"H":100}
+    hp["name_map"] = {"H":"thickness"}
+    hp["source"] = "mat"
+    hp["X_map"] = {"x1":"x", "x2":"y"}
+    hp["shapefile"] = os.path.join(repoPath, "dataset", expFileName)
+
+    d = pinn.domain.Domain( pinn.parameter.DomainParameter(hp))
+    p = SingleDataParameter(hp)
+    data_loader = MatData(p)
+    data_loader.load_data(d)
     data_loader.prepare_training_data()
-    assert(data_loader.X['H'].shape[1] == 1)
+    assert(data_loader.X['H'].shape == (100,2))
+    assert(data_loader.sol['H'].shape == (100,1))
+    icoord = data_loader.get_ice_coordinates()
+    assert icoord.shape == (673,2)
 
-    hp["X_map"] = {"x":"t", "y":"y"}
+def test_MatData_physics():
+    filename = "flightTracks.mat"
+    expFileName = "fastflow_CF.exp"
+    repoPath = os.path.dirname(__file__) + "/../examples/"
+    appDataPath = os.path.join(repoPath, "dataset")
+    path = os.path.join(appDataPath, filename)
+
+    hp = {}
+    hp["data_path"] = path
+    hp["data_size"] = {"H":100}
+    hp["name_map"] = {"H":"thickness"}
+    hp["source"] = "mat"
+    hp["X_map"] = {"x1":"x", "x2":"y"}
+    hp["shapefile"] = os.path.join(repoPath, "dataset", expFileName)
+    hp["equations"] = {"SSA":{"input":["x1", "x2"]}}
+
+    phy = pinn.physics.Physics(pinn.parameter.PhysicsParameter(hp))
     p = SingleDataParameter(hp)
     data_loader = MatData(p)
-    data_loader.load_data()
-    assert ("y" in data_loader.X_dict)
-    assert ("x" not in data_loader.X_dict)
+    data_loader.load_data(physics=phy)
+    data_loader.prepare_training_data()
+    assert(data_loader.X['H'].shape == (100,2))
+    assert(data_loader.sol['H'].shape == (100,1))
+    icoord = data_loader.get_ice_coordinates()
+    assert icoord.shape == (3192,2)
 
 def test_h5Data():
     filename = "subdomain_data.h5"