Fixed a huge issue where base loss centroid convergence was not using 'one' as its reference loss function, resulting in opposite trends.

analysis/base_loss/centroid_convergence_plotter.py

@@ -178,9 +178,9 @@ def process_condition(condition_name):
             continue
         final_idx = data_dict["epochs"].index(final_epoch)
         final_theta = data_dict["theta_over_epochs"][final_idx]
-        final_loss_val = replicate_base_loss(final_theta, desired_theta, base_key)
+        final_loss_val = replicate_base_loss(final_theta, desired_theta, "one")
 
-        conv_epoch = find_convergence_epoch(base_key, data_dict, desired_theta, threshold)
+        conv_epoch = find_convergence_epoch("one", data_dict, desired_theta, threshold)
         exponent_val = base_loss_functions[base_key][0]  # e.g. 1/5, 3, 4, etc.
 
         results.append({

analysis/base_loss_learning_rate_sweep/centroid_convergence_plotter.py

@@ -178,9 +178,9 @@ def process_condition(condition_name):
             continue
         final_idx = data_dict["epochs"].index(final_epoch)
         final_theta = data_dict["theta_over_epochs"][final_idx]
-        final_loss_val = replicate_base_loss(final_theta, desired_theta, base_key)
+        final_loss_val = replicate_base_loss(final_theta, desired_theta, "one")
 
-        conv_epoch = find_convergence_epoch(base_key, data_dict, desired_theta, threshold)
+        conv_epoch = find_convergence_epoch("one", data_dict, desired_theta, threshold)
         exponent_val = base_loss_functions[base_key][0]  # e.g. 1/5, 3, 4, etc.
 
         results.append({

analysis/misc/validation/controller_1000/in-sample/summary.csv

@@ -0,0 +1,2 @@
+Run #,theta0,omega0,alpha0,desired_theta,final_theta,loss
+1,12.566370614359172,0.0,0.0,0.0,-0.006831930121820209,4717.5843021979645

analysis/misc/validation/controller_1000/out-of-sample/summary.csv

@@ -0,0 +1 @@
+Run #,theta0,omega0,alpha0,desired_theta,final_theta,loss

analysis/misc/validator.py

@@ -3,19 +3,18 @@ import torch.nn as nn
 from torchdiffeq import odeint
 import numpy as np
 import os
-import shutil
-import csv
-import math
+
 import matplotlib.pyplot as plt
 import pandas as pd
 
+import sys
+sys.path.append("/home/judson/Neural-Networks-in-GNC/inverted_pendulum/analysis")
 from PendulumController import PendulumController
 
 # List of controller file names to validate.
 # Replace these paths with your actual controller file paths.
 controller_file_names = [
-    "/home/judson/Neural-Networks-in-GNC/inverted_pendulum/training/base_loss_learning_rate_sweep/one_fourth/lr_0.300/controllers/controller_199.pth",
-    "/home/judson/Neural-Networks-in-GNC/inverted_pendulum/training/base_loss_learning_rate_sweep/four/lr_0.200/controllers/controller_200.pth"
+    "/home/judson/Neural-Networks-in-GNC/inverted_pendulum/training/time_weighting/constant/controllers/controller_1000.pth",
 ]
 
 # Constants for simulation
@@ -30,51 +29,51 @@ t_eval = np.linspace(0, T, num_steps)
 # In-sample validation cases: [theta0, omega0, alpha0, desired_theta]
 in_sample_cases = [
     (1/6 * np.pi,    0.0, 0.0, 0.0),
-    (-1/6 * np.pi,   0.0, 0.0, 0.0),
-    (2/3 * np.pi,    0.0, 0.0, 0.0),
-    (-2/3 * np.pi,   0.0, 0.0, 0.0),
-    (0.0, 1/3 * np.pi,      0.0, 0.0),
-    (0.0, -1/3 * np.pi,     0.0, 0.0),
-    (0.0, 2 * np.pi,        0.0, 0.0),
-    (0.0, -2 * np.pi,       0.0, 0.0),
-    (0.0, 0.0, 0.0,     2*np.pi),
-    (0.0, 0.0, 0.0,     -2*np.pi),
-    (0.0, 0.0, 0.0,     1/2 * np.pi),
-    (0.0, 0.0, 0.0,     -1/2 * np.pi),
-    (0.0, 0.0, 0.0,     1/3 * np.pi),
-    (0.0, 0.0, 0.0,     -1/3 * np.pi),
-    (1/4 * np.pi,    1 * np.pi, 0.0, 0.0),
-    (-1/4 * np.pi,   -1 * np.pi, 0.0, 0.0),
-    (1/2 * np.pi,    -1 * np.pi, 0.0, 1/3 * np.pi),
-    (-1/2 * np.pi,   1 * np.pi, 0.0, -1/3 * np.pi),
-    (1/4 * np.pi,    1 * np.pi, 0.0, 2 * np.pi),
-    (-1/4 * np.pi,   -1 * np.pi, 0.0, 2 * np.pi),
-    (1/2 * np.pi,    -1 * np.pi, 0.0, 4 * np.pi),
-    (-1/2 * np.pi,   1 * np.pi, 0.0, -4 * np.pi),
+    # (-1/6 * np.pi,   0.0, 0.0, 0.0),
+    # (2/3 * np.pi,    0.0, 0.0, 0.0),
+    # (-2/3 * np.pi,   0.0, 0.0, 0.0),
+    # (0.0, 1/3 * np.pi,      0.0, 0.0),
+    # (0.0, -1/3 * np.pi,     0.0, 0.0),
+    # (0.0, 2 * np.pi,        0.0, 0.0),
+    # (0.0, -2 * np.pi,       0.0, 0.0),
+    # (0.0, 0.0, 0.0,     2*np.pi),
+    # (0.0, 0.0, 0.0,     -2*np.pi),
+    # (0.0, 0.0, 0.0,     1/2 * np.pi),
+    # (0.0, 0.0, 0.0,     -1/2 * np.pi),
+    # (0.0, 0.0, 0.0,     1/3 * np.pi),
+    # (0.0, 0.0, 0.0,     -1/3 * np.pi),
+    # (1/4 * np.pi,    1 * np.pi, 0.0, 0.0),
+    # (-1/4 * np.pi,   -1 * np.pi, 0.0, 0.0),
+    # (1/2 * np.pi,    -1 * np.pi, 0.0, 1/3 * np.pi),
+    # (-1/2 * np.pi,   1 * np.pi, 0.0, -1/3 * np.pi),
+    # (1/4 * np.pi,    1 * np.pi, 0.0, 2 * np.pi),
+    # (-1/4 * np.pi,   -1 * np.pi, 0.0, 2 * np.pi),
+    # (1/2 * np.pi,    -1 * np.pi, 0.0, 4 * np.pi),
+    # (-1/2 * np.pi,   1 * np.pi, 0.0, -4 * np.pi),
 ]
 
 # Out-of-sample validation cases (generated previously)
 out_sample_cases = [
-    (-2.198958, -4.428501, 0.450833, 0.000000),
-    (1.714196, -0.769896, 0.202738, 0.000000),
-    (0.241195, -5.493715, 0.438996, 0.000000),
-    (0.030605, 4.901513, -0.479243, 0.000000),
-    (1.930445, -1.301926, -0.454050, 0.000000),
-    (-0.676063, 4.246865, 0.036303, 0.000000),
-    (0.734920, -5.925202, 0.047097, 0.000000),
-    (-3.074471, -3.535424, 0.315438, 0.000000),
-    (-0.094486, 6.111091, 0.150525, 0.000000),
-    (-1.647671, 5.720526, 0.334181, 0.000000),
-    (-2.611260, 5.087704, 0.045460, -3.610785),
-    (1.654137, 0.982081, -0.192725, 1.003872),
-    (-2.394899, 3.550547, -0.430938, 3.261897),
-    (0.474917, 0.555166, -0.285173, 1.866752),
-    (-0.640369, -4.678490, -0.340663, 3.150098),
-    (1.747517, -3.248204, -0.001520, 1.221787),
-    (2.505283, -2.875006, -0.065617, -3.690269),
-    (1.337244, 2.221707, 0.044979, -2.459730),
-    (1.531012, 2.230981, -0.291206, -1.924535),
-    (-1.065792, 4.320740, 0.075405, -1.550644),
+    # (-2.198958, -4.428501, 0.450833, 0.000000),
+    # (1.714196, -0.769896, 0.202738, 0.000000),
+    # (0.241195, -5.493715, 0.438996, 0.000000),
+    # (0.030605, 4.901513, -0.479243, 0.000000),
+    # (1.930445, -1.301926, -0.454050, 0.000000),
+    # (-0.676063, 4.246865, 0.036303, 0.000000),
+    # (0.734920, -5.925202, 0.047097, 0.000000),
+    # (-3.074471, -3.535424, 0.315438, 0.000000),
+    # (-0.094486, 6.111091, 0.150525, 0.000000),
+    # (-1.647671, 5.720526, 0.334181, 0.000000),
+    # (-2.611260, 5.087704, 0.045460, -3.610785),
+    # (1.654137, 0.982081, -0.192725, 1.003872),
+    # (-2.394899, 3.550547, -0.430938, 3.261897),
+    # (0.474917, 0.555166, -0.285173, 1.866752),
+    # (-0.640369, -4.678490, -0.340663, 3.150098),
+    # (1.747517, -3.248204, -0.001520, 1.221787),
+    # (2.505283, -2.875006, -0.065617, -3.690269),
+    # (1.337244, 2.221707, 0.044979, -2.459730),
+    # (1.531012, 2.230981, -0.291206, -1.924535),
+    # (-1.065792, 4.320740, 0.075405, -1.550644),
 ]
 
 # Define the ODE integration step using RK4, returning the new state and the computed torque.