Merge pull request #14 from jakekrell/main

v3.1.0

Examples/GP Integrate/GP_integrate_example.py

@@ -2,79 +2,82 @@
 from FoKL.GP_Integrate import GP_Integrate
 import numpy as np
 import matplotlib.pyplot as plt
-import warnings
-warnings.filterwarnings("ignore", category=UserWarning)
 
 
-# ======================================================================================================================
+def main():
 
-# Inputs:
-traininputs = np.loadtxt('traininputs.txt',dtype=float,delimiter=',')
-traindata1 = np.loadtxt('traindata1.txt',dtype=float,delimiter=',')
-traindata2 = np.loadtxt('traindata2.txt',dtype=float,delimiter=',')
-traindata = [traindata1, traindata2]
-y = np.loadtxt('y.txt',dtype=float,delimiter=',')
-utest = np.loadtxt('utest.csv',dtype=float,delimiter=',')
+    # Inputs:
+    traininputs = np.loadtxt('traininputs.txt', dtype=float, delimiter=',')
+    traindata1 = np.loadtxt('traindata1.txt', dtype=float, delimiter=',')
+    traindata2 = np.loadtxt('traindata2.txt', dtype=float, delimiter=',')
+    traindata = [traindata1, traindata2]
+    y = np.loadtxt('y.txt', dtype=float, delimiter=',')
+    utest = np.loadtxt('utest.csv', dtype=float, delimiter=',')
 
-# User-defined constant hyperparameters (to override default values):
-relats_in = [1,1,1,1,1,1]
-a = 1000
-b = 1
-draws = 2000
-way3 = True
-threshav = 0
-threshstda = 0
-threshstdb = 100
+    # User-defined hyperparameters (to override default values):
+    relats_in = [1, 1, 1, 1, 1, 1]
+    a = 1000
+    b = 1
+    draws = 2000
+    way3 = True
+    threshav = 0
+    threshstda = 0
+    threshstdb = 100
 
-# Initializing FoKL model with constant hypers:
-model = FoKLRoutines.FoKL(relats_in=relats_in, a=a, b=b, draws=draws, way3=way3, threshav=threshav, threshstda=threshstda, threshstdb=threshstdb)
+    # Initializing FoKL model with constant hypers:
+    model = FoKLRoutines.FoKL(relats_in=relats_in, a=a, b=b, draws=draws, way3=way3, threshav=threshav,
+                              threshstda=threshstda, threshstdb=threshstdb)
 
-# User-defined variable hyperparameters (to iterate through either for different data or for sweeping the same data):
-btau = [0.6091, 1]
+    # User-defined hyperparameters (to iterate through for different data or to sweep the same data):
+    btau = [0.6091, 1]
 
-# Iterating through datasets:
-betas = []
-mtx = []
-for ii in range(2):
+    # Iterating through datasets:
 
-    print("\nCurrently fitting model to dataset", int(ii+1),". . .")
+    betas = []
+    mtx = []
+    for ii in range(2):
 
-    # Updating model with current iteration of variable hyperparameters:
-    model.btau = btau[ii]
+        print("\nCurrently fitting model to dataset", int(ii+1), "...")
 
-    # Running emulator routine for current model/data:
-    betas_i, mtx_i, _ = model.fit(traininputs, traindata[ii])
+        # Updating model with current iteration of variable hyperparameters:
+        model.btau = btau[ii]
 
-    print("Done!")
+        # Running emulator routine for current model/data:
+        betas_i, mtx_i, _ = model.fit(traininputs, traindata[ii])
 
-    # Store values for post-processing (i.e., GP integration):
-    betas.append(betas_i[1000:])
-    mtx.append(mtx_i)
+        print("Done!")
 
-    # Clear all attributes (except for hypers) so previous results do not influence the next iteration:
-    model.clear()
+        # Store values for post-processing (i.e., GP integration):
+        betas.append(betas_i[1000:])
+        mtx.append(mtx_i)
 
-# ======================================================================================================================
+        # Clear all attributes (except for hypers) so previous results do not influence the next iteration:
+        model.clear()
 
-# Integrating with FoKL.GP_Integrate():
+    # Integrating with FoKL.GP_Integrate():
 
-phis = model.phis # same for all models iterated through, so just grab value from most recent model
+    phis = model.phis # same for all models iterated through, so just grab value from most recent model
 
-n,m = np.shape(y)
-norms1 = [np.min(y[0,0:int(m/2)]),np.max(y[0,0:int(m/2)])]
-norms2 = [np.min(y[1,0:int(m/2)]),np.max(y[1,0:int(m/2)])]
-norms = np.transpose([norms1,norms2])
+    n, m = np.shape(y)
+    norms1 = [np.min(y[0, 0:int(m/2)]), np.max(y[0, 0:int(m/2)])]
+    norms2 = [np.min(y[1, 0:int(m/2)]), np.max(y[1, 0:int(m/2)])]
+    norms = np.transpose([norms1, norms2])
 
-start = 4
-stop = 3750*4
-stepsize = 4
-used_inputs = [[1,1,1],[1,1,1]]
-ic = y[:,int(m/2)-1]
+    start = 4
+    stop = 3750*4
+    stepsize = 4
+    used_inputs = [[1, 1, 1], [1, 1, 1]]
+    ic = y[:, int(m/2)-1]
 
-T, Y = GP_Integrate([np.mean(betas[0],axis=0),np.mean(betas[1],axis=0)], [mtx[0],mtx[1]], utest, norms, phis, start, stop, ic, stepsize, used_inputs)
+    T, Y = GP_Integrate([np.mean(betas[0], axis=0), np.mean(betas[1], axis=0)], [mtx[0],mtx[1]], utest, norms, phis,
+                        start, stop, ic, stepsize, used_inputs)
 
-plt.figure()
-plt.plot(T,Y[0],T,y[0][3750:7500])
-plt.plot(T,Y[1],T,y[1][3750:7500])
-plt.show()
+    plt.figure()
+    plt.plot(T, Y[0], T, y[0][3750:7500])
+    plt.plot(T, Y[1], T, y[1][3750:7500])
+    plt.show()
+
+
+if __name__ == '__main__':
+    main()
 

Examples/Sigmoid/SigmoidTest.py

@@ -1,64 +1,64 @@
 from FoKL import FoKLRoutines
 import numpy as np
-import warnings
-warnings.filterwarnings("ignore", category=UserWarning)
 
 
-# ======================================================================================================================
+def main():
 
-# Inputs:
-X_grid = np.loadtxt('X.csv',dtype=float,delimiter=',')
-Y_grid = np.loadtxt('Y.csv',dtype=float,delimiter=',')
+    # Inputs:
+    X_grid = np.loadtxt('X.csv',dtype=float,delimiter=',')
+    Y_grid = np.loadtxt('Y.csv',dtype=float,delimiter=',')
 
-# Data:
-Z_grid = np.loadtxt('DATA_nois.csv',dtype=float,delimiter=',')
+    # Data:
+    Z_grid = np.loadtxt('DATA_nois.csv',dtype=float,delimiter=',')
 
-# Reshaping grid matrices into vectors via fortran index order:
-m, n = np.shape(X_grid) # = np.shape(Y_grid) = np.shape(Z_grid) = dimensions of grid
-X = np.reshape(X_grid, (m*n,1), order='F')
-Y = np.reshape(Y_grid, (m*n,1), order='F')
-Z = np.reshape(Z_grid, (m*n,1), order='F')
+    # Reshaping grid matrices into vectors via fortran index order:
+    m, n = np.shape(X_grid) # = np.shape(Y_grid) = np.shape(Z_grid) = dimensions of grid
+    X = np.reshape(X_grid, (m*n,1), order='F')
+    Y = np.reshape(Y_grid, (m*n,1), order='F')
+    Z = np.reshape(Z_grid, (m*n,1), order='F')
 
-# Initializing FoKL model with some user-defined hyperparameters (leaving others blank for default values):
-model = FoKLRoutines.FoKL(a=9, b=0.01, atau=3, btau=4000, aic=True)
+    # Initializing FoKL model with some user-defined hyperparameters (leaving others blank for default values):
+    model = FoKLRoutines.FoKL(a=9, b=0.01, atau=3, btau=4000, aic=True)
 
-# Training FoKL model on a random selection of 100% (or 100%, 80%, 60%, etc.) of the dataset:
-train_all = [1] # = [1, 0.8, 0.6] etc. if sweeping through the percentage of data to train on
-betas_all = []
-mtx_all = []
-evs_all = []
-meen_all = []
-bounds_all = []
-rmse_all = []
-for train in train_all:
+    # Training FoKL model on a random selection of 100% (or 100%, 80%, 60%, etc.) of the dataset:
+    train_all = [1] # = [1, 0.8, 0.6] etc. if sweeping through the percentage of data to train on
+    betas_all = []
+    mtx_all = []
+    evs_all = []
+    meen_all = []
+    bounds_all = []
+    rmse_all = []
+    for train in train_all:
 
-    print("\nCurrently fitting model to",train * 100,"% of data . . .")
+        print("\nCurrently fitting model to",train * 100,"% of data ...")
 
-    # Running emulator routine to fit model to training data as a function of the corresponding training inputs:
-    betas, mtx, evs = model.fit([X, Y], Z, train=train)
+        # Running emulator routine to fit model to training data as a function of the corresponding training inputs:
+        betas, mtx, evs = model.fit([X, Y], Z, train=train)
 
-    # Provide feedback to user before the figure from coverage3() pops up and pauses the code:
-    print("\nDone! Please close the figure to continue.\n")
+        # Provide feedback to user before the figure from coverage3() pops up and pauses the code:
+        print("\nDone! Please close the figure to continue.\n")
 
-    # Evaluating and visualizing predicted values of data as a function of all inputs (train set plus test set):
-    title = 'FoKL Model Trained on ' + str(train * 100) + '% of Data'
-    meen, bounds, rmse = model.coverage3(plot='bounds',title=title,legend=1)
+        # Evaluating and visualizing predicted values of data as a function of all inputs (train set plus test set):
+        title = 'FoKL Model Trained on ' + str(train * 100) + '% of Data'
+        meen, bounds, rmse = model.coverage3(plot='bounds',title=title,legend=1)
 
-    # Store any values from iteration if performing additional post-processing or analysis:
-    betas_all.append(betas)
-    mtx_all.append(mtx)
-    evs_all.append(evs)
-    meen_all.append(meen)
-    bounds_all.append(bounds)
-    rmse_all.append(rmse)
+        # Store any values from iteration if performing additional post-processing or analysis:
+        betas_all.append(betas)
+        mtx_all.append(mtx)
+        evs_all.append(evs)
+        meen_all.append(meen)
+        bounds_all.append(bounds)
+        rmse_all.append(rmse)
 
-    # Reset the model so that all attributes of the FoKL class are removed except for the hyperparameters:
-    model.clear()
+        # Reset the model so that all attributes of the FoKL class are removed except for the hyperparameters:
+        model.clear()
 
-# ======================================================================================================================
+    # Post-processing:
+    print("\nThe results are as follows:")
+    for ii in range(len(train_all)):
+        print("\n   ",train_all[ii]*100,"% of Data:\n    --> RMSE =",rmse_all[ii])
 
-# Post-processing:
-print("\nThe results are as follows:")
-for ii in range(len(train_all)):
-    print("\n   ",train_all[ii]*100,"% of Data:\n    --> RMSE =",rmse_all[ii])
+
+if __name__ == '__main__':
+    main()