Moved dA calculation in initialization

kandread · Oct 15, 2019 · b4418f0 · b4418f0
1 parent 7f70563
commit b4418f0
Showing 1 changed file with 9 additions and 12 deletions.
diff --git a/confluence/integrator.py b/confluence/integrator.py
@@ -30,6 +30,11 @@ def __init__(self, n, A0, H, W, S, dA, routing_table):
         self.nreaches = len(A0)
         self.A0 = np.array(A0)
         self.n = np.array(n)
+        w = np.mean(W, axis=0)
+        h = np.mean(H, axis=0)
+        if dA is None:
+            dA = np.array([(w[r] + W[np.argmin(H[:, r]), r]) / 2 * (h[r] - H[np.argmin(H[:, r]), r])
+                           for r in range(self.nreaches)]).T
         self.data = (H, W, S, dA)
         self.rivs = self._riverTopology(routing_table)
 
@@ -55,12 +60,8 @@ def objective(self, x):
         H, W, S, dA = self.data
         w = np.mean(W, axis=0)
         h = np.mean(H, axis=0)
-        if dA is None:
-            dA = np.array([(w[r] + W[np.argmin(H[:, r]), r]) / 2 * (h[r] - H[np.argmin(H[:, r]), r])
-                           for r in range(self.nreaches)]).T
-        else:
-            dA = np.mean(dA, axis=0)
-        Q = 1 / n * (A0 + dA)**(5 / 3) * w**(-2 / 3) * np.mean(S,
+        da = np.mean(dA, axis=0)
+        Q = 1 / n * (A0 + da)**(5 / 3) * w**(-2 / 3) * np.mean(S,
                                                                axis=0)**(1 / 2)
         error = [np.sqrt((Q[i] - np.sum(Q[j] for j in self.rivs[i]))**2) for i in self.rivs]
         return np.sum(error)
@@ -72,12 +73,8 @@ def constraint(self, i, x):
         H, W, S, dA = self.data
         w = np.mean(W, axis=0)
         h = np.mean(H, axis=0)
-        if dA is None:
-            dA = np.array([(w[r] + W[np.argmin(H[:, r]), r]) / 2 * (h[r] - H[np.argmin(H[:, r]), r])
-                           for r in range(self.nreaches)]).T
-        else:
-            dA = np.mean(dA, axis=0)
-        Q = 1 / n * (A0 + dA)**(5 / 3) * w**(-2 / 3) * np.mean(S,
+        da = np.mean(dA, axis=0)
+        Q = 1 / n * (A0 + da)**(5 / 3) * w**(-2 / 3) * np.mean(S,
                                                                axis=0)**(1 / 2)
         return Q[i] - np.sum([Q[j] for j in self.rivs[i]])