File size: 6,283 Bytes
d2b8aad | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 | """Visualization utilities for physics-informed Bayesian optimization."""
from typing import Callable, Dict, List, Optional, Tuple
import torch
from torch import Tensor
import numpy as np
def plot_convergence(
campaign_df,
maximize: bool = True,
title: str = "Optimization Convergence",
figsize: Tuple[int, int] = (10, 6),
):
"""Plot the optimization convergence curve.
Args:
campaign_df: DataFrame from OptimizationCampaign.to_dataframe().
maximize: Whether the objective is being maximized.
title: Plot title.
figsize: Figure size.
Returns:
matplotlib Figure.
"""
import matplotlib.pyplot as plt
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=figsize)
objectives = campaign_df["objective"].values
# Left: all observations
ax1.plot(range(len(objectives)), objectives, "o-", alpha=0.6, markersize=4)
ax1.set_xlabel("Experiment Number")
ax1.set_ylabel("Objective")
ax1.set_title("All Observations")
ax1.grid(True, alpha=0.3)
# Right: best-so-far
if maximize:
best_so_far = np.maximum.accumulate(objectives)
else:
best_so_far = np.minimum.accumulate(objectives)
ax2.plot(range(len(best_so_far)), best_so_far, "s-", color="green", markersize=4)
ax2.set_xlabel("Experiment Number")
ax2.set_ylabel("Best Objective")
ax2.set_title("Best So Far")
ax2.grid(True, alpha=0.3)
fig.suptitle(title, fontsize=14)
plt.tight_layout()
return fig
def plot_surrogate_1d(
surrogate,
bounds: Tuple[float, float],
X_observed: Optional[Tensor] = None,
y_observed: Optional[Tensor] = None,
physics_fn: Optional[Callable] = None,
true_fn: Optional[Callable] = None,
n_grid: int = 200,
title: str = "Surrogate Model",
figsize: Tuple[int, int] = (10, 6),
):
"""Plot a 1D surrogate model with confidence intervals.
Args:
surrogate: A SurrogateModel instance.
bounds: (lower, upper) for the 1D input.
X_observed: Observed inputs (n, 1).
y_observed: Observed outputs (n, 1).
physics_fn: Optional physics model for comparison.
true_fn: Optional true function for comparison.
n_grid: Number of grid points.
title: Plot title.
figsize: Figure size.
Returns:
matplotlib Figure.
"""
import matplotlib.pyplot as plt
fig, ax = plt.subplots(figsize=figsize)
X_grid = torch.linspace(bounds[0], bounds[1], n_grid).unsqueeze(-1).to(torch.float64)
mean, var = surrogate.predict(X_grid)
std = var.sqrt()
x_np = X_grid.squeeze().numpy()
mean_np = mean.squeeze().detach().numpy()
std_np = std.squeeze().detach().numpy()
# Surrogate prediction
ax.plot(x_np, mean_np, "b-", label="Surrogate Mean", linewidth=2)
ax.fill_between(
x_np,
mean_np - 2 * std_np,
mean_np + 2 * std_np,
alpha=0.2,
color="blue",
label="95% CI",
)
# Physics model
if physics_fn is not None:
with torch.no_grad():
physics_pred = physics_fn(X_grid).squeeze().numpy()
ax.plot(x_np, physics_pred, "r--", label="Physics Model", linewidth=1.5)
# True function
if true_fn is not None:
with torch.no_grad():
true_pred = true_fn(X_grid).squeeze().numpy()
ax.plot(x_np, true_pred, "k-", label="True Function", linewidth=1.5, alpha=0.7)
# Observations
if X_observed is not None and y_observed is not None:
ax.scatter(
X_observed.squeeze().numpy(),
y_observed.squeeze().numpy(),
c="red",
s=50,
zorder=5,
label="Observations",
edgecolors="black",
)
ax.set_xlabel("Input")
ax.set_ylabel("Output")
ax.set_title(title)
ax.legend()
ax.grid(True, alpha=0.3)
plt.tight_layout()
return fig
def plot_surrogate_2d(
surrogate,
bounds: Tensor,
param_names: Tuple[str, str] = ("x1", "x2"),
X_observed: Optional[Tensor] = None,
n_grid: int = 50,
title: str = "Surrogate Model (2D)",
figsize: Tuple[int, int] = (12, 5),
):
"""Plot 2D surrogate model as contour plots (mean and uncertainty).
Args:
surrogate: A SurrogateModel instance.
bounds: Tensor of shape (2, 2) with [lower, upper] bounds.
param_names: Names of the two parameters.
X_observed: Observed inputs (n, 2).
n_grid: Grid resolution per dimension.
title: Plot title.
figsize: Figure size.
Returns:
matplotlib Figure.
"""
import matplotlib.pyplot as plt
x1 = torch.linspace(float(bounds[0, 0]), float(bounds[1, 0]), n_grid)
x2 = torch.linspace(float(bounds[0, 1]), float(bounds[1, 1]), n_grid)
X1, X2 = torch.meshgrid(x1, x2, indexing="ij")
X_grid = torch.stack([X1.flatten(), X2.flatten()], dim=-1).to(torch.float64)
mean, var = surrogate.predict(X_grid)
mean_2d = mean.squeeze().reshape(n_grid, n_grid).detach().numpy()
std_2d = var.sqrt().squeeze().reshape(n_grid, n_grid).detach().numpy()
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=figsize)
# Mean
c1 = ax1.contourf(
X1.numpy(), X2.numpy(), mean_2d, levels=20, cmap="viridis"
)
plt.colorbar(c1, ax=ax1)
ax1.set_xlabel(param_names[0])
ax1.set_ylabel(param_names[1])
ax1.set_title("Predicted Mean")
# Uncertainty
c2 = ax2.contourf(
X1.numpy(), X2.numpy(), std_2d, levels=20, cmap="plasma"
)
plt.colorbar(c2, ax=ax2)
ax2.set_xlabel(param_names[0])
ax2.set_ylabel(param_names[1])
ax2.set_title("Predicted Std Dev")
# Overlay observations
if X_observed is not None:
for ax in [ax1, ax2]:
ax.scatter(
X_observed[:, 0].numpy(),
X_observed[:, 1].numpy(),
c="red",
s=30,
edgecolors="white",
zorder=5,
)
fig.suptitle(title, fontsize=14)
plt.tight_layout()
return fig
|