100 Lens Sample with Substructure

complexLA/Analysis/metrics.py

@@ -0,0 +1,257 @@
+import numpy as np
+from statistics import median
+
+#def single function for general case of n situations
+def PerturberSample(sample_num, y_test, y_pred_wop, y_pred_wp, y_pred_wpl, std_pred_wop, std_pred_wp, std_pred_wpl):
+    """ Calculuate the median error, median absolute error, and the median standard deviation for the Wop, WP, and WPL cases.
+
+    Parameters
+    ----------
+    sample_num : int
+        Number of mock sample lenses.
+    y_test : np.array
+        Known values of the lens parameters.
+    y_pred_wop, y_pred_wp, y_pred_wpl : np.array
+        Predicted values of the lens parameters for the wop, wp, and wpl cases.
+    std_pred_wop, std_pred_wp, std_pred_wpl : np.array
+        Standard deviations of the lens parameter predicted distributions.
+
+    Returns
+    -------
+
+    """
+
+    ##ME - Median Error
+    param_diff=[]
+    for i in range(sample_num-1):
+        param_diff.append(y_pred_wop[i]-y_test[i])
+    #print(param_diff)
+    sum=0
+    for i in param_diff:
+        #print(i)
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    ME_wop = list(np.round(mean,2))
+    #print(ME_wop)
+
+    param_diff=[]
+    for i in range(sample_num-1):
+        param_diff.append(y_pred_wp[i]-y_test[i])
+    sum=0
+    for i in param_diff:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    ME_wp = list(np.round(mean,2))
+    #print(ME_wp)
+
+    param_diff=[]
+    for i in range(sample_num-1):
+        param_diff.append(y_pred_wpl[i]-y_test[i])
+    sum=0
+    for i in param_diff:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    ME_wpl = list(np.round(mean,2))
+    #print(ME_wpl)
+
+    ##MAE - Median Absolute Error
+    param_diff_abs=[]
+    for i in range(sample_num-1):
+        param_diff_abs.append(abs(y_pred_wop[i]-y_test[i]))
+    #print(param_diff_abs)
+    sum=0
+    for i in param_diff_abs:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    MAE_wop = list(np.round(mean,2))
+    #print(MAE_wop)
+
+    param_diff_abs=[]
+    for i in range(sample_num-1):
+        param_diff_abs.append(abs(y_pred_wp[i]-y_test[i]))
+    sum=0
+    for i in param_diff_abs:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    MAE_wp = list(np.round(mean,2))
+    #print(MAE_wp)
+
+    param_diff_abs=[]
+    for i in range(sample_num-1):
+        param_diff_abs.append(abs(y_pred_wpl[i]-y_test[i]))
+    sum=0
+    for i in param_diff_abs:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    MAE_wpl = list(np.round(mean,2))
+    #print(MAE_wpl)
+
+    ##Msigma - Median Standard Deviation
+    #print(std_pred_wop)
+    sum=0
+    for i in std_pred_wop:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    Mstd_wop = list(np.round(mean,2))
+    #print(Mstd_wop)
+
+    ##Msigma - Median Standard Deviation
+    sum=0
+    for i in std_pred_wp:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    Mstd_wp = list(np.round(mean,2))
+    #print(Mstd_wp)
+
+    ##Msigma - Median Standard Deviation
+    sum=0
+    for i in std_pred_wpl:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    Mstd_wpl = list(np.round(mean,2))
+    #print(Mstd_wpl)
+
+    metrics = [ME_wop, ME_wp, ME_wpl, MAE_wop, MAE_wp, MAE_wpl, Mstd_wop, Mstd_wp, Mstd_wpl]
+
+    return metrics
+
+def PerturberSampleTrunc(sample_num, y_test, y_pred_wop, y_pred_wp, y_pred_wpl, std_pred_wop, std_pred_wp, std_pred_wpl):
+
+    #WoP
+    param_diff=[]
+    for i in range(sample_num-1):
+        param_diff.append(y_pred_wop[i]-y_test[i])
+    #print(param_diff)
+
+    ##ME - Mean Error
+    sum=0
+    for i in param_diff:
+        #print(i)
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    ME_wop = list(np.round(mean,2))
+    #print(ME_wop)
+
+    ##MDE - Median Error
+    #for i in range(param_num):
+     #   param_i = []
+    #for row in param_diff:
+     #   for i in range(param_num):
+      #      param_i.append(row[i])
+    #for i in param_num: 
+     #   print(param_i)
+
+
+    #Median_wop = list(np.round(median(param_diff),2))
+
+     #WP
+    param_diff=[]
+    for i in range(sample_num-1):
+        param_diff.append(y_pred_wp[i]-y_test[i])
+    #print(param_diff)
+    ##ME - Mean Error
+    sum=0
+    for i in param_diff:
+        #print(i)
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    ME_wp = list(np.round(mean,2))
+
+    #WPL
+    param_diff=[]
+    for i in range(sample_num-1):
+        param_diff.append(y_pred_wpl[i]-y_test[i])
+    ##ME - Mean Error
+    sum=0
+    for i in param_diff:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    ME_wpl = list(np.round(mean,2))
+    #print(ME_wpl)
+    ##MDE - Median Error
+    #Median_wpl = list(np.round(median(param_diff),2))
+
+    ##Msigma - Mean Standard Deviation
+    sum=0
+    for i in std_pred_wop:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    Mstd_wop = list(np.round(mean,2))
+    #print(Mstd_wop)
+    ##MDsigma - Median Standard Deviation
+    #MDstd_wop = list(np.round(median(std_pred_wop),2))
+
+    ##Msigma - Mean Standard Deviation
+    sum=0
+    for i in std_pred_wp:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    Mstd_wp = list(np.round(mean,2))
+    #print(Mstd_wp)
+    ##MDsigma - Median Standard Deviation
+    #MDstd_wp = list(np.round(median(std_pred_wp),2))
+
+    ##Msigma - Median Standard Deviation
+    sum=0
+    for i in std_pred_wpl:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    Mstd_wpl = list(np.round(mean,2))
+    #print(Mstd_wpl)
+    ##MDsigma - Median Standard Deviation
+    #MDstd_wpl = list(np.round(median(std_pred_wpl),2))
+
+    mean_metrics = [ME_wop, ME_wp, ME_wpl, Mstd_wop, Mstd_wp, Mstd_wpl]
+    #median_metrics = [Median_wop, Median_wpl, MDstd_wop, MDstd_wpl]
+
+    return mean_metrics
+
+def PerturberSampleTrunc_Substructure(sample_num, y_test, y_pred_WoS, y_pred_WS, std_pred_WoS, std_pred_WS):
+
+    #WoS
+    param_diff=[]
+    for i in range(sample_num-1):
+        param_diff.append(y_pred_WoS[i]-y_test[i])
+    #print(param_diff)
+
+    ##ME - Mean Error
+    sum=0
+    for i in param_diff:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    ME_WoS = list(np.round(mean,2))
+    #print(ME_wop)
+
+     #WS
+    param_diff=[]
+    for i in range(sample_num-1):
+        param_diff.append(y_pred_WS[i]-y_test[i])
+    #print(param_diff)
+
+    ##ME - Mean Error
+    sum=0
+    for i in param_diff:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    ME_WS = list(np.round(mean,2))
+
+    ##Msigma - Mean Standard Deviation
+    sum=0
+    for i in std_pred_WoS:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    Mstd_WoS = list(np.round(mean,2))
+    #print(Mstd_wop)
+
+    ##Msigma - Mean Standard Deviation
+    sum=0
+    for i in std_pred_WS:
+        sum=sum+i
+    mean=(sum/(sample_num-1))
+    Mstd_WS = list(np.round(mean,2))
+    #print(Mstd_wp)
+
+    mean_metrics = [ME_WoS, ME_WS, Mstd_WoS, Mstd_WS]
+    #median_metrics = [Median_wop, Median_wpl, MDstd_wop, MDstd_wpl]
+
+    return mean_metrics

complexLA/Analysis/network_predictions.py

@@ -0,0 +1,59 @@
+import os
+import sys
+from deep_lens_modeling import network_predictions
+import time
+
+#define a function for a general case of n situations
+def Predictions(config_file, image_folder, sample_num=1):
+    """ Generates network predictions from images
+
+    Parameters
+    ----------
+    config_file : str
+        File location for the config file being used.
+    image_path : str
+        Image folder location where the images will be stored.
+    sample_num : int
+        Number of mock sample lenses. Default 1
+
+    Returns
+    -------
+    y_test_wop : float
+        Truth values of lens parameters.
+    y_pred_wop, y_pred_wp, y_pred_wpl : float
+        Prediction values for the wop, wp, and wpl cases.
+    std_pred_wop, std_pred_wp, std_pred_wpl : float
+        Standard deviation of the prediction values for the wop, wp and wpl cases.
+    prec_pred_wop, prec_pred_wp, prec_pred_wpl : array
+        Predicted precision matrix for the wop, wp and wpl cases.
+
+    """
+
+    #Generating the images
+    import subprocess
+    command = f'python /Users/Logan/AppData/Local/Programs/Python/Python312/Lib/site-packages/paltas/paltas/generate.py {config_file} {image_folder} --n {sample_num} --tf_record'
+    print(f"Running: {command}")
+    result = subprocess.run(command, shell=True, capture_output=True, text=True)
+
+    print("STDOUT:", result.stdout)
+    print("STDERR:", result.stderr)
+    print("Exit code:", result.returncode)
+    #os.system(f'python /Users/Logan/AppData/Local/Programs/Python/Python312/Lib/site-packages/paltas/paltas/generate.py {config_file} {image_folder} --n {sample_num-1} --tf_record')
+
+    # Compute predictions for test sets
+    path_to_weights = '../docs/xresnet34_068--14.58.h5'
+    path_to_norms = '../docs/norms.csv'
+
+    learning_params = ['main_deflector_parameters_theta_E','main_deflector_parameters_gamma1',
+                       'main_deflector_parameters_gamma2','main_deflector_parameters_gamma',
+                       'main_deflector_parameters_e1','main_deflector_parameters_e2',
+                       'main_deflector_parameters_center_x','main_deflector_parameters_center_y',
+                       'source_parameters_center_x','source_parameters_center_y']
+
+    model_predictions = network_predictions.NetworkPredictions(path_to_weights,path_to_norms,
+                                                               learning_params,loss_type='diag',model_type='xresnet34',norm_type='lognorm')
+
+    # generate network predictions
+    y_test, y_pred, std_pred, prec_pred = model_predictions.gen_network_predictions(image_folder)
+
+    return y_test, y_pred, std_pred, prec_pred