File size: 5,397 Bytes
69bf174
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
"""

Basic example demonstrating CHG algorithm usage

"""

import numpy as np
import matplotlib.pyplot as plt
from chg_algorithm import CHG, CHGOptimizer


def basic_regression_example():
    """Demonstrate basic regression with CHG"""
    print("=== Basic CHG Regression Example ===")
    
    # Generate synthetic 1D data for visualization
    np.random.seed(42)
    X_train = np.random.uniform(-3, 3, (50, 1))
    y_train = np.sin(X_train.flatten()) + 0.1 * np.random.randn(50)
    
    X_test = np.linspace(-4, 4, 100).reshape(-1, 1)
    y_true = np.sin(X_test.flatten())
    
    # Initialize and fit CHG model
    model = CHG(input_dim=1, hidden_dim=16, num_heads=2)
    pred_mean, pred_var = model.fit_predict(X_train, y_train, X_test)
    pred_std = np.sqrt(pred_var)
    
    # Print metrics
    mse = np.mean((pred_mean - y_true)**2)
    print(f"Mean Squared Error: {mse:.4f}")
    print(f"Log Marginal Likelihood: {model.log_marginal_likelihood(X_train, y_train):.4f}")
    
    # Visualization
    plt.figure(figsize=(10, 6))
    plt.scatter(X_train.flatten(), y_train, alpha=0.6, label='Training Data')
    plt.plot(X_test.flatten(), y_true, 'r-', label='True Function')
    plt.plot(X_test.flatten(), pred_mean, 'b-', label='CHG Prediction')
    plt.fill_between(X_test.flatten(), 
                     pred_mean - 2*pred_std, 
                     pred_mean + 2*pred_std, 
                     alpha=0.2, label='95% Confidence')
    plt.xlabel('Input')
    plt.ylabel('Output')
    plt.title('CHG Gaussian Process Regression')
    plt.legend()
    plt.grid(True, alpha=0.3)
    plt.show()


def optimization_example():
    """Demonstrate parameter optimization"""
    print("\n=== CHG Optimization Example ===")
    
    # Generate data
    np.random.seed(123)
    X_train = np.random.randn(80, 2)
    y_train = np.sum(X_train**2, axis=1) + 0.5 * np.random.randn(80)
    
    # Initialize model and optimizer
    model = CHG(input_dim=2, hidden_dim=20, num_heads=3)
    optimizer = CHGOptimizer(model, learning_rate=0.01)
    
    # Track optimization progress
    lml_history = []
    
    print("Optimizing CHG parameters...")
    for epoch in range(20):
        optimizer.step(X_train, y_train)
        lml = model.log_marginal_likelihood(X_train, y_train)
        lml_history.append(lml)
        
        if epoch % 5 == 0:
            print(f"Epoch {epoch:2d}: Log Marginal Likelihood = {lml:.4f}")
    
    # Plot optimization progress
    plt.figure(figsize=(8, 5))
    plt.plot(lml_history, 'b-', linewidth=2)
    plt.xlabel('Epoch')
    plt.ylabel('Log Marginal Likelihood')
    plt.title('CHG Optimization Progress')
    plt.grid(True, alpha=0.3)
    plt.show()
    
    print(f"Final Log Marginal Likelihood: {lml_history[-1]:.4f}")


def uncertainty_quantification_example():
    """Demonstrate uncertainty quantification capabilities"""
    print("\n=== Uncertainty Quantification Example ===")
    
    # Generate noisy data with outliers
    np.random.seed(456)
    X_train = np.random.uniform(-2, 2, (60, 1))
    y_clean = 0.5 * X_train.flatten()**3 - X_train.flatten()
    
    # Add noise and some outliers
    noise = 0.2 * np.random.randn(60)
    outlier_idx = np.random.choice(60, 5, replace=False)
    noise[outlier_idx] += np.random.choice([-2, 2], 5) * 2  # Add outliers
    
    y_train = y_clean + noise
    
    X_test = np.linspace(-3, 3, 80).reshape(-1, 1)
    
    # Fit CHG model
    model = CHG(input_dim=1, hidden_dim=12, num_heads=2)
    pred_mean, pred_var = model.fit_predict(X_train, y_train, X_test)
    pred_std = np.sqrt(pred_var)
    
    # Analyze uncertainties
    high_uncertainty_idx = pred_std > np.percentile(pred_std, 75)
    print(f"Percentage of high-uncertainty predictions: {np.mean(high_uncertainty_idx)*100:.1f}%")
    print(f"Average prediction uncertainty: {np.mean(pred_std):.4f}")
    print(f"Maximum prediction uncertainty: {np.max(pred_std):.4f}")
    
    # Visualization
    plt.figure(figsize=(12, 5))
    
    plt.subplot(1, 2, 1)
    plt.scatter(X_train.flatten(), y_train, alpha=0.7, c='red', label='Training Data (with outliers)')
    plt.plot(X_test.flatten(), pred_mean, 'b-', linewidth=2, label='CHG Prediction')
    plt.fill_between(X_test.flatten(), 
                     pred_mean - 2*pred_std, 
                     pred_mean + 2*pred_std, 
                     alpha=0.3, label='95% Confidence')
    plt.xlabel('Input')
    plt.ylabel('Output')
    plt.title('CHG Predictions with Uncertainty')
    plt.legend()
    plt.grid(True, alpha=0.3)
    
    plt.subplot(1, 2, 2)
    plt.plot(X_test.flatten(), pred_std, 'g-', linewidth=2)
    plt.fill_between(X_test.flatten()[high_uncertainty_idx], 
                     0, pred_std[high_uncertainty_idx], 
                     alpha=0.4, color='orange', 
                     label='High Uncertainty Regions')
    plt.xlabel('Input')
    plt.ylabel('Prediction Uncertainty (σ)')
    plt.title('Uncertainty Estimation')
    plt.legend()
    plt.grid(True, alpha=0.3)
    
    plt.tight_layout()
    plt.show()


if __name__ == "__main__":
    # Run all examples
    basic_regression_example()
    optimization_example()
    uncertainty_quantification_example()
    
    print("\n=== All Examples Completed ===")