Reduce function building overhead in wrappedCircuit

README.md

@@ -35,6 +35,8 @@ impedance.py requires:
 -   Matplotlib (>=3.5)
 -   Altair (>=3.0)
 -   Pandas
+-   pygad>=3.6.0
+-   pyswarms>=1.3
 
 Several example notebooks are provided in the `docs/source/examples/` directory. Opening these will require Jupyter notebook or Jupyter lab.
 

docs/source/validation.rst

@@ -1,7 +1,7 @@
 Validation
 ==========
 
-Interpreting EIS data fundamentally relies on the the system conforming to
+Interpreting EIS data fundamentally relies on the system conforming to
 conditions of causality, linearity, and stability. For an example of how
 the adherence to the Kramers-Kronig relations, see the `Validation Example Jupyter Notebook
 <examples/validation_example.ipynb>`_

impedance/models/circuits/circuits.py

@@ -134,20 +134,14 @@ def predict(self, frequencies, use_initial=False):
             Predicted impedance at each frequency
         """
         frequencies = np.array(frequencies, dtype=float)
-
+        buildCircuit_text=buildCircuit(self.circuit, constants=self.constants, 
+                            eval_string='', index=0)[0]
+        builtCircuit = eval('lambda frequencies,parameters : ' +  buildCircuit_text, circuit_elements)
         if self._is_fit() and not use_initial:
-            return eval(buildCircuit(self.circuit, frequencies,
-                                     *self.parameters_,
-                                     constants=self.constants, eval_string='',
-                                     index=0)[0],
-                        circuit_elements)
+            return builtCircuit(frequencies,self.parameters_)
         else:
             warnings.warn("Simulating circuit based on initial parameters")
-            return eval(buildCircuit(self.circuit, frequencies,
-                                     *self.initial_guess,
-                                     constants=self.constants, eval_string='',
-                                     index=0)[0],
-                        circuit_elements)
+            return builtCircuit(frequencies,self.initial_guess)
 
     def get_param_names(self):
         """ Converts circuit string to names and units """

impedance/models/circuits/elements.py

@@ -166,7 +166,14 @@ def Wo(p, f):
     """
     omega = 2 * np.pi * np.array(f)
     Z0, tau = p[0], p[1]
-    Z = Z0 / (np.sqrt(1j * omega * tau) * np.tanh(np.sqrt(1j * omega * tau)))
+
+    arg = np.sqrt(1j * omega * tau)
+    tanh = np.ones_like(arg, dtype=complex)
+    tanh[arg.real <= -100] = -1.0
+    mask = np.abs(arg.real) < 100
+    tanh[mask] = np.tanh(arg[mask])
+
+    Z = Z0 / (arg * tanh)
     return Z  # Zw(omega)
 
 
@@ -186,7 +193,14 @@ def Ws(p, f):
     """
     omega = 2 * np.pi * np.array(f)
     Z0, tau = p[0], p[1]
-    Z = Z0 * np.tanh(np.sqrt(1j * omega * tau)) / np.sqrt(1j * omega * tau)
+
+    arg = np.sqrt(1j * omega * tau)
+    tanh = np.ones_like(arg, dtype=complex)
+    tanh[arg.real <= -100] = -1.0
+    mask = np.abs(arg.real) < 100
+    tanh[mask] = np.tanh(arg[mask])
+
+    Z = Z0 * tanh / arg
     return Z
 
 
@@ -288,10 +302,14 @@ def Gs(p, f):
     """
     omega = 2 * np.pi * np.array(f)
     R_G, t_G, phi = p[0], p[1], p[2]
-    Z = R_G / (
-        np.sqrt(1 + 1j * omega * t_G)
-        * np.tanh(phi * np.sqrt(1 + 1j * omega * t_G))
-    )
+
+    arg = phi * np.sqrt(1 + 1j * omega * t_G)
+    tanh = np.ones_like(arg, dtype=complex)
+    tanh[arg.real <= -100] = -1.0
+    mask = np.abs(arg.real) < 100
+    tanh[mask] = np.tanh(arg[mask])
+
+    Z = R_G / (np.sqrt(1 + 1j * omega * t_G) * tanh)
     return Z
 
 
@@ -350,7 +368,14 @@ def TLMQ(p, f):
     omega = 2 * np.pi * np.array(f)
     Rion, Qs, gamma = p[0], p[1], p[2]
     Zs = 1 / (Qs * (1j * omega) ** gamma)
-    Z = np.sqrt(Rion * Zs) / np.tanh(np.sqrt(Rion / Zs))
+
+    arg = np.sqrt(Rion / Zs)
+    tanh = np.ones_like(arg, dtype=complex)
+    tanh[arg.real <= -100] = -1.0
+    mask = np.abs(arg.real) < 100
+    tanh[mask] = np.tanh(arg[mask])
+
+    Z = np.sqrt(Rion * Zs) / tanh
     return Z
 
 
@@ -388,14 +413,17 @@ def T(p, f):
     A, B, a, b = p[0], p[1], p[2], p[3]
     beta = (a + 1j * omega * b) ** (1 / 2)
 
-    sinh = []
-    for x in beta:
-        if x < 100:
-            sinh.append(np.sinh(x))
-        else:
-            sinh.append(1e10)
+    mask = np.abs(beta.real) < 100
+
+    sinh = np.ones_like(beta, dtype=complex) * 1e10
+    sinh[beta.real <= -100] = -1e10
+    sinh[mask] = np.sinh(beta[mask])
+
+    tanh = np.ones_like(beta, dtype=complex)
+    tanh[beta.real <= -100] = -1.0
+    tanh[mask] = np.tanh(beta[mask])
 
-    Z = A / (beta * np.tanh(beta)) + B / (beta * np.array(sinh))
+    Z = A / (beta * tanh) + B / (beta * sinh)
     return Z