From d38ce3e1d1e3d27531d96245567c07f824482894 Mon Sep 17 00:00:00 2001 From: Vats Date: Wed, 25 Mar 2026 16:35:10 +0100 Subject: [PATCH] GP evaluation --- .vscode/settings.json | 1 + .../crest_heights_north_sea/GP_eval.ipynb | 1325 +++++++++++++++++ 2 files changed, 1326 insertions(+) create mode 100644 examples/crest_heights_north_sea/GP_eval.ipynb diff --git a/.vscode/settings.json b/.vscode/settings.json index 86e54207..0be023e1 100644 --- a/.vscode/settings.json +++ b/.vscode/settings.json @@ -49,4 +49,5 @@ "mypy-type-checker.reportingScope": "workspace", "mypy-type-checker.preferDaemon": true, "ruff.configurationPreference": "filesystemFirst", + "python-envs.pythonProjects": [], } \ No newline at end of file diff --git a/examples/crest_heights_north_sea/GP_eval.ipynb b/examples/crest_heights_north_sea/GP_eval.ipynb new file mode 100644 index 00000000..05dc2280 --- /dev/null +++ b/examples/crest_heights_north_sea/GP_eval.ipynb @@ -0,0 +1,1325 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 44, + "metadata": {}, + "outputs": [], + "source": [ + "from __future__ import annotations\n", + "\n", + "import sys\n", + "import warnings\n", + "from pathlib import Path\n", + "from typing import Any, NamedTuple\n", + "\n", + "import numpy as np # type: ignore[import-not-found]\n", + "import torch\n", + "from ax import SearchSpace\n", + "from ax.core import RangeParameter\n", + "from botorch.fit import fit_gpytorch_mll\n", + "from botorch.models import SingleTaskGP\n", + "from botorch.models.transforms import Normalize\n", + "from gpytorch.mlls import ExactMarginalLogLikelihood\n", + "from matplotlib import pyplot as plt\n", + "from scipy import stats\n", + "\n", + "torch.set_default_dtype(torch.float64)\n", + "\n", + "# Add notebook directory to sys.path so local modules (e.g. problem.py) can be imported.\n", + "_examples_dir = Path.cwd()\n", + "if str(_examples_dir) not in sys.path:\n", + " sys.path.insert(0, str(_examples_dir))\n", + "\n", + "from problem import SEARCH_SPACE, make_exp # type: ignore[import-not-found] # noqa: E402\n" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Extracting Training Data as Tensors\n", + "\n", + "We could have called model.train_inputs to get the Ax DataFrame directly, but **Ax applies internal transforms** so train_inputs is already normalised and standardised across all points.\n", + "\n", + "We do not want that, normalising/standardising based on all-points statistics leaks information about the held-out point. For proper LOO cross-validation, each fold must compute its own normalisation and standardisation statistics from the n−1 training points only.\n", + "\n", + "The function below reads the Ax experiment's DataFrame and assembles (X, Y, Yvar) in the original problem space (untransformed).\n", + "\n", + "\n", + "### Observation Noise Variance (Yvar)\n", + "\n", + "Yvar is the observation noise variance derived from the Gumbel MLE estimate:\n", + "$$\\text{sem} = \\sqrt{\\text{Var}[\\hat{\\theta}_\\text{MLE}]} \\quad \\Rightarrow \\quad Y_\\text{var} = \\text{sem}^2$$\n", + "\n", + "This captures the aleatoric (label) noise/finite-sample estimation error from a limited number of simulator draws per point (e.g., 100).\n", + "\n", + "### Ax DataFrame Structure\n", + "\n", + "Ax links input parameters to output metrics via arms:\n", + "\n", + "experiment.trials\n", + "└── Trial (index 0)\n", + " └── Arm \"0_0\"\n", + " ├── arm.parameters = {\"Hs\": 4.2, \"Tp\": 8.7} → X\n", + " └── trial.run_metadata[\"simulation_result\"]\n", + " → LocalMetadataMetric reads it\n", + " → df row: arm_name=\"0_0\", metric_name=\"loc\"\n", + " mean=12.3, sem=0.4 # mean → Y, sem² → Yvar" + ] + }, + { + "cell_type": "code", + "execution_count": 45, + "metadata": {}, + "outputs": [], + "source": [ + "def extract_training_tensors( # noqa: C901\n", + " experiment: Any, # noqa: ANN401\n", + " metric_names: list[str],\n", + " parameter_names: list[str],\n", + ") -> tuple[torch.Tensor, torch.Tensor, torch.Tensor]:\n", + " \"\"\"Extract training data from an Ax experiment into tensors.\n", + "\n", + " Args:\n", + " experiment: The finished Ax Experiment.\n", + " metric_names: Ordered list of output metric names, e.g. [\"loc\", \"scale\"].\n", + " parameter_names: Ordered list of input parameter names, e.g. [\"Hs\", \"Tp\"].\n", + "\n", + " Returns:\n", + " X: (n, d) input locations in problem space.\n", + " Y: (n, m) output means (GP training targets, in problem space).\n", + " Yvar: (n, m) output variance = sem^2 (aleatoric/observation noise, heteroscedastic).\n", + "\n", + " \"\"\"\n", + " # Collects results from completed trials as (arm_name, metric_name, mean, sem) rows.\n", + " df = experiment.fetch_data().df\n", + " print(f\"Extracted raw data frame with {len(df)} rows from Ax experiment.\")\n", + " print(f\"Available metrics: {df['metric_name'].unique()}\")\n", + " print(f\"A row example: {df}\")\n", + "\n", + " df = df[df[\"metric_name\"].isin(metric_names)].copy()\n", + "\n", + " # Build a parameter lookup: arm_name -> parameter dict.\n", + " arm_params: dict[str, dict] = {}\n", + " for trial in experiment.trials.values():\n", + " for arm in trial.arms:\n", + " arm_params[arm.name] = arm.parameters\n", + "\n", + " arm_names_ordered: list[str] = []\n", + " seen: set[str] = set()\n", + " for trial in sorted(experiment.trials.values(), key=lambda t: t.index):\n", + " for arm in trial.arms:\n", + " if arm.name not in seen:\n", + " arm_names_ordered.append(arm.name)\n", + " seen.add(arm.name)\n", + "\n", + " xs, ys, yvars = [], [], []\n", + " for arm_name in arm_names_ordered:\n", + " arm_df = df[df[\"arm_name\"] == arm_name]\n", + " if arm_df.empty:\n", + " continue\n", + "\n", + " params = arm_params[arm_name]\n", + " x = [params[p] for p in parameter_names]\n", + "\n", + " y_row, yvar_row = [], []\n", + " for mn in metric_names:\n", + " row = arm_df[arm_df[\"metric_name\"] == mn]\n", + " if row.empty:\n", + " continue\n", + " mean_val = float(row[\"mean\"].values[0]) # Gumbel MLE\n", + " sem_val = float(row[\"sem\"].values[0]) # sqrt of Fisher covariance\n", + " y_row.append(mean_val)\n", + " yvar_row.append(sem_val**2) # variance = sem², the aleatoric/observation noise\n", + "\n", + " if len(y_row) == len(metric_names):\n", + " xs.append(x)\n", + " ys.append(y_row)\n", + " yvars.append(yvar_row)\n", + "\n", + " X = torch.tensor(xs, dtype=torch.float64) # (n, d) # noqa: N806\n", + " Y = torch.tensor(ys, dtype=torch.float64) # (n, m) # noqa: N806\n", + " Yvar = torch.tensor(yvars, dtype=torch.float64) # (n, m) # noqa: N806\n", + "\n", + " return X, Y, Yvar\n", + "\n", + "\n", + "def make_bounds_tensor(search_space: SearchSpace, parameter_names: list[str]) -> torch.Tensor:\n", + " \"\"\"Return (2, d) bounds tensor [[lower, ...], [upper, ...]] from an Ax SearchSpace.\n", + "\n", + " This is used to build the Normalize input transform, matching Ax's UnitX transform.\n", + " \"\"\"\n", + " lowers, uppers = [], []\n", + " for name in parameter_names:\n", + " p = search_space.parameters[name]\n", + " assert isinstance(p, RangeParameter), f\"Only RangeParameters supported, got {type(p)}\"\n", + " lowers.append(p.lower)\n", + " uppers.append(p.upper)\n", + "\n", + " return torch.tensor([lowers, uppers], dtype=torch.float64) # (2, d)\n" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Building the GP : Applied Transforms\n", + "\n", + "Since we imported untransformed tensors, we manually apply the same transforms that Ax/BoTorch uses internally when fitting the GP:\n", + "\n", + "| Transform | Does |\n", + "|-----------|---------|\n", + "| NormalizeX | Normalises input X to $[0, 1]$ using the search space bounds |\n", + "| StandardizeY | Standardises each output metric to zero mean and unit std,built into SingleTaskGP by default |\n", + "| Kernel | Matérn 5/2 (BoTorch default) |\n", + "\n", + "\n", + "### Assumption: Fixed Observation Noise\n", + "\n", + "Since train_Yvar is provided, BoTorch uses [FixedNoiseGaussianLikelihood](see here : https://botorch.readthedocs.io/en/v0.15.1/_modules/botorch/models/gp_regression.html). The sem² values in train_Yvar are treated as ground-truth observation noise, i.e., they are not learnt or adjusted during hyperparameter optimisation." + ] + }, + { + "cell_type": "code", + "execution_count": 46, + "metadata": {}, + "outputs": [], + "source": [ + "def build_gp(\n", + " train_X: torch.Tensor, # noqa: N803\n", + " train_Y: torch.Tensor, # noqa: N803\n", + " train_Yvar: torch.Tensor, # noqa: N803\n", + " bounds: torch.Tensor,\n", + " *,\n", + " fit: bool = True,\n", + ") -> SingleTaskGP:\n", + " \"\"\"Build and fit a SingleTaskGP on the given training data.\n", + "\n", + " Args:\n", + " train_X: (n, d) tensor of input locations in problem space.\n", + " train_Y: (n, m) tensor of output values in problem space.\n", + " train_Yvar: (n, m) tensor of output variances in problem space.\n", + " bounds: (2, d) tensor of input bounds.\n", + " fit: Whether to fit the GP. Set False to skip fitting (e.g. for debugging).\n", + "\n", + " Returns:\n", + " A fitted (or unfitted if fit=False) SingleTaskGP model.\n", + "\n", + " \"\"\"\n", + " d = train_X.shape[-1]\n", + " input_transform = Normalize(d=d, bounds=bounds)\n", + "\n", + " model = SingleTaskGP(\n", + " train_X=train_X,\n", + " train_Y=train_Y,\n", + " train_Yvar=train_Yvar,\n", + " input_transform=input_transform,\n", + " # StandardizeY is applied by default inside SingleTaskGP.\n", + " )\n", + "\n", + " if fit:\n", + " mll = ExactMarginalLogLikelihood(model.likelihood, model)\n", + " with warnings.catch_warnings():\n", + " warnings.simplefilter(\"ignore\")\n", + " fit_gpytorch_mll(mll) # L-BFGS-B; maximises the marginal log-likelihood\n", + "\n", + " model.eval()\n", + " return model\n", + "\n", + "\n", + "class LOOFoldResult(NamedTuple):\n", + " \"\"\"Result of one Leave-One-Out fold, with one point left out.\"\"\"\n", + "\n", + " idx: int # index of the left-out point\n", + " x_left_out: torch.Tensor # (d,) left-out input location in problem space\n", + " y_left_out: torch.Tensor # (m,) left-out output value (mean) in problem space\n", + " yvar_left_out: torch.Tensor # (m,) left-out output variance (sem²) in problem space\n", + " y_pred_mean: torch.Tensor # (m,) predicted mean at left-out point (GP on n-1 points)\n", + " y_pred_var_epistemic: torch.Tensor # (m,) GP epistemic variance only (no observation noise)\n", + " pred_var_total: torch.Tensor # (m,) total predictive variance = epistemic + aleatoric\n" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Running the Leave-One-Out Validation Loop\n", + "\n", + "For each point $i$ in the training data:\n", + "\n", + "1. Build the GP on $(X_{-i},\\, Y_{-i},\\, Y_\\text{var,-i})$, all points except $i$.\n", + "2. Fit kernel hyperparameters (Matérn length-scale and output scale) via MLL on these $n-1$ points.\n", + "3. Compute the **epistemic posterior** at point $i$ (observation_noise=False) : uncertainty due to lack of data.\n", + "4. Compute the **total predictive posterior** at point $i$ (observation_noise=True) : epistemic + aleatoric uncertainty.\n", + "\n", + "#### Why both Step 3 and Step 4?\n", + "\n", + "- **Step 3** isolates the epistemic uncertainty at the held-out point, which reflects the GP's ability to generalise to unseen locations. This is the primary signal for LOO evaluation.\n", + "- **Step 4** additionally accounts for aleatoric uncertainty (observation noise). Near a training point the epistemic variance would be close to zero, but if the observation noise is high (e.g. because only a small number of simulations were used to fit the Gumbel parameters), that uncertainty still matters and should be included in the predictive distribution.\n", + "- Together, Step 4 answers: *\"Is the GP uncertain because the observation itself is noisy?\"*\n", + "\n", + "#### Important Note: \n", + "How does observation_noise=True work for an unseen point?\n", + "Sample noise is a property of the input data only, so how does the model estimate aleatoric uncertainty for a held-out point it has never seen?\n", + "\n", + "- When SingleTaskGP is initialised with train_Yvar, BoTorch uses a FixedNoiseGaussianLikelihood. For an unseen test point, it simply returns mean(Yvar): the mean of the training noise values rather than a point-specific estimate. (See botorch/models/gpytorch.py, line 408.)\n", + "- The BoTorch developers note this is a simplification. The underlying assumption is that label noise does not vary dramatically across input space. This is not a severe limitation, but worth being aware of: our noise is heteroscedastic, and without additional assumptions there is no better way to estimate it at an unseen location.\n", + "- As the number of simulator runs per point increases, the observation noise Yvar decreases and becomes better controlled.\n", + "- Consequence for this script: yvar_left_out is the exact (ground-truth) noise at the held-out point, whereas pred_var_total − pred_var_epistemic is BoTorch's estimate i.e., the mean of the $n-1$ remaining training noise values." + ] + }, + { + "cell_type": "code", + "execution_count": 47, + "metadata": {}, + "outputs": [], + "source": [ + "def run_loo(\n", + " X: torch.Tensor, # noqa: N803\n", + " Y: torch.Tensor, # noqa: N803\n", + " Yvar: torch.Tensor, # noqa: N803\n", + " bounds: torch.Tensor,\n", + " indices: list[int] | None = None,\n", + " *,\n", + " verbose: bool = True,\n", + ") -> list[LOOFoldResult]:\n", + " \"\"\"Run Leave-One-Out cross validation on the given training data.\n", + "\n", + " Args:\n", + " X: (n, d) tensor of input locations in problem space.\n", + " Y: (n, m) tensor of output values in problem space.\n", + " Yvar: (n, m) tensor of output variances in problem space.\n", + " bounds: (2, d) tensor of input bounds for building the GP.\n", + " indices: Optional list of indices to run LOO on. If None, runs on all points.\n", + " verbose: Print info per fold.\n", + "\n", + " Returns:\n", + " A list of LOOFoldResult, one for each fold.\n", + " \"\"\"\n", + " n = X.shape[0]\n", + " if indices is None:\n", + " indices = list(range(n))\n", + "\n", + " results: list[LOOFoldResult] = []\n", + "\n", + " for fold, i in enumerate(indices):\n", + " if verbose:\n", + " print(f\"\\nRunning LOO fold {fold + 1}/{len(indices)}: leaving out point index {i}...\")\n", + "\n", + " keep = [j for j in range(n) if j != i]\n", + " X_loo = X[keep] # (n-1, d) # noqa: N806\n", + " Y_loo = Y[keep] # (n-1, m) # noqa: N806\n", + " Yvar_loo = Yvar[keep] # (n-1, m) # noqa: N806\n", + "\n", + " gp_model = build_gp(X_loo, Y_loo, Yvar_loo, bounds, fit=True)\n", + "\n", + " x_left_out = X[i].unsqueeze(0) # (1, d)\n", + "\n", + " with torch.no_grad():\n", + " # Epistemic posterior: uncertainty from not having seen this point.\n", + " posterior_epistemic = gp_model.posterior(x_left_out, observation_noise=False)\n", + " posterior_total = gp_model.posterior(x_left_out, observation_noise=True)\n", + "\n", + " pred_mean = posterior_total.mean.squeeze(0) # (m,)\n", + " pred_var_epistemic = posterior_epistemic.variance.squeeze(0) # (m,)\n", + " pred_var_total = posterior_total.variance.squeeze(0) # (m,)\n", + "\n", + " results.append(\n", + " LOOFoldResult(\n", + " idx=i,\n", + " x_left_out=X[i],\n", + " y_left_out=Y[i],\n", + " yvar_left_out=Yvar[i],\n", + " y_pred_mean=pred_mean,\n", + " y_pred_var_epistemic=pred_var_epistemic,\n", + " pred_var_total=pred_var_total,\n", + " )\n", + " )\n", + "\n", + " if verbose:\n", + " print(f\"Fold {fold + 1}/{len(indices)}: Left out point index {i}, x={x_left_out.tolist()}\")\n", + " print(f\" True y={Y[i].tolist()}, sem²={Yvar[i].tolist()}\")\n", + " print(f\" Predicted mean={pred_mean.tolist()}\")\n", + " print(f\" Predicted epistemic var={pred_var_epistemic.tolist()}\")\n", + " print(f\" Predicted total var={pred_var_total.tolist()}\")\n", + " return results\n", + "\n", + "\n", + "class LOOMetrics(NamedTuple):\n", + " \"\"\"Metrics computed from the LOO results.\"\"\"\n", + "\n", + " metric_name: str\n", + " n_folds: int\n", + "\n", + " mae: float # mean absolute error\n", + " rmse: float # root mean squared error\n", + "\n", + " # Uncertainty calibration metrics:\n", + " z_scores: np.ndarray # type: ignore # (n,) (y_pred_mean - y_left_out) / sqrt(pred_var_total)\n", + " z_mean: float # mean of z-scores, should be close to 0\n", + " z_std: float # std of z-scores, should be close to 1\n", + " ci_95_coverage: float # fraction inside predicted 95% CI (should be ~0.95)\n", + "\n", + " frac_aleatoric: float # mean(aleatoric / pred_var_total) across held-out points\n", + "\n", + "\n", + "def compute_metrics(loo_results: list[LOOFoldResult], metric_idx: int, metric_name: str) -> LOOMetrics:\n", + " \"\"\"Compute LOO metrics for a specific output metric (e.g., loc or scale).\n", + "\n", + " Args:\n", + " loo_results: List of LOOFoldResult per fold.\n", + " metric_idx: Index of the output metric (e.g., 0 for loc, 1 for scale).\n", + " metric_name: Name of the metric.\n", + "\n", + " Returns:\n", + " LOOMetrics containing the computed metrics.\n", + " \"\"\"\n", + " y_true = np.array([r.y_left_out[metric_idx].item() for r in loo_results])\n", + " y_pred = np.array([r.y_pred_mean[metric_idx].item() for r in loo_results])\n", + " pred_var_total = np.array([r.pred_var_total[metric_idx].item() for r in loo_results])\n", + " true_var_alea = np.array([r.yvar_left_out[metric_idx].item() for r in loo_results])\n", + " pred_var_epi = np.array([r.y_pred_var_epistemic[metric_idx].item() for r in loo_results])\n", + " pred_var_alea = pred_var_total - pred_var_epi\n", + "\n", + " # Clamp for numerical stability in z-score computation.\n", + " pred_var_total = np.maximum(pred_var_total, 1e-12)\n", + "\n", + " residuals = y_pred - y_true\n", + " mae = np.mean(np.abs(residuals))\n", + " rmse = np.sqrt(np.mean(residuals**2))\n", + "\n", + " z_scores = residuals / np.sqrt(pred_var_total)\n", + " z_mean = np.mean(z_scores)\n", + " z_std = np.std(z_scores)\n", + " coverage_95 = np.mean(np.abs(z_scores) <= 1.96) # noqa: PLR2004\n", + "\n", + " # Using estimated aleatoric (pred_var_alea) for consistency with what the GP uses internally.\n", + " frac_aleatoric = np.mean(pred_var_alea / pred_var_total)\n", + "\n", + " return LOOMetrics(\n", + " metric_name=metric_name,\n", + " n_folds=len(loo_results),\n", + " mae=mae,\n", + " rmse=rmse,\n", + " z_scores=z_scores,\n", + " z_mean=z_mean,\n", + " z_std=z_std,\n", + " ci_95_coverage=coverage_95,\n", + " frac_aleatoric=frac_aleatoric,\n", + " )\n" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Plots\n", + "\n", + "**1. Z-score histogram with standard normal overlay**: overlap indicates good calibration of GP uncertainties.\n", + "- Mean close to 0, variance close to 1.\n", + "- Tails fatter than normal: occasionally large errors, due to kernel choice or poor fit.\n", + "- Too spiky: z-score variance is low, GP variance is too large.\n", + "\n", + "**2. Q-Q plot of z-scores against a standard normal distribution**: shows tails more clearly than the histogram.\n", + "- Points approximately on the line $y = x$: uncertainties well calibrated.\n", + "- Line shifted or curved, or tails deviating: bias, miscalibrated uncertainties, or outliers.\n", + "\n", + "**3. Predicted vs Actual plot** with error bars showing predictive uncertainty intervals (95% CI).\n", + "- Points on the diagonal indicate accurate predictions.\n", + "- Error bars show $\\sqrt{\\text{total\\_var}}$ = predictive standard deviation (epistemic + aleatoric).\n", + "\n", + "**4. Per-point epistemic vs aleatoric uncertainty bar plot.**\n", + "- At each point, how much of the uncertainty is epistemic (lack of data) vs aleatoric (inherent observation noise).\n", + "- Indicates where adding more data would help (high epistemic) vs where label noise needs to be managed (high aleatoric)." + ] + }, + { + "cell_type": "code", + "execution_count": 48, + "metadata": {}, + "outputs": [], + "source": [ + "def plot_loo_results( # noqa: PLR0915\n", + " loo_results: list[LOOFoldResult],\n", + " metrics_list: list[LOOMetrics],\n", + " metric_names: list[str],\n", + ") -> plt.Figure:\n", + " \"\"\"Plot LOO cross-validation results for each output metric.\n", + "\n", + " Args:\n", + " loo_results: List of LOOFoldResult per fold.\n", + " metrics_list: List of LOOMetrics per metric.\n", + " metric_names: List of metric display names.\n", + "\n", + " Returns:\n", + " A matplotlib Figure containing the plots.\n", + " \"\"\"\n", + " m = len(metrics_list)\n", + " n_rows = 4\n", + "\n", + " fig, axes = plt.subplots(n_rows, m, figsize=(6 * m, 4 * n_rows))\n", + "\n", + " if m == 1:\n", + " axes = axes.reshape(n_rows, 1)\n", + "\n", + " for col, (loom, display_name) in enumerate(zip(metrics_list, metric_names, strict=True)):\n", + " metric_idx = col\n", + " z_scores = loom.z_scores\n", + " y_true = np.array([r.y_left_out[metric_idx].item() for r in loo_results])\n", + " y_pred = np.array([r.y_pred_mean[metric_idx].item() for r in loo_results])\n", + " pred_var_total = np.array([r.pred_var_total[metric_idx].item() for r in loo_results])\n", + " std_total = np.sqrt(pred_var_total)\n", + " true_var_alea = np.array([r.yvar_left_out[metric_idx].item() for r in loo_results])\n", + " pred_var_epi = np.array([r.y_pred_var_epistemic[metric_idx].item() for r in loo_results])\n", + " pred_var_alea = pred_var_total - pred_var_epi\n", + "\n", + " # 1. Z-score histogram with standard normal overlay\n", + " ax = axes[0, col]\n", + " ax.hist(z_scores, bins=10, density=True, alpha=0.6, color=\"g\", label=\"LOO z-scores\")\n", + " xr = np.linspace(-5, 5, 100)\n", + " ax.plot(xr, stats.norm.pdf(xr), \"r-\", lw=2, label=\"N(0,1) reference\")\n", + " ax.axvline(0, color=\"grey\", linestyle=\"--\", lw=1)\n", + " ax.set_xlabel(\"Z-score\")\n", + " ax.set_ylabel(\"Density\")\n", + " ax.set_title(\n", + " f\"{display_name}\\nZ-score histogram\\n\"\n", + " f\"mean={loom.z_mean:+.3f}, std={loom.z_std:.3f} (ideal: mean=0, std=1)\"\n", + " )\n", + " ax.legend()\n", + "\n", + " # 2. Q-Q plot of z-scores against standard normal\n", + " ax = axes[1, col]\n", + " (osm, osr), (slope, intercept, r) = stats.probplot(z_scores, dist=\"norm\")\n", + " ax.scatter(osm, osr, color=\"blue\", label=f\"R={r:.3f}\")\n", + " ax.plot(osm, float(slope) * np.array(osm) + float(intercept), \"r-\", lw=2, label=\"Fit\")\n", + " ax.set_title(f\"Q-Q Plot: {display_name}\")\n", + " ax.set_xlabel(\"Normal Quantiles\")\n", + " ax.set_ylabel(\"z-score Quantiles\")\n", + " ax.legend()\n", + "\n", + " # 3. Predicted vs Actual with 95% CI error bars\n", + " ax = axes[2, col]\n", + " ax.errorbar(\n", + " y_true,\n", + " y_pred,\n", + " yerr=1.96 * np.sqrt(pred_var_epi), # so 1.96 * std_total 95% CI (epistemic + aleatoric) can do yerr=1.96*sqrt(pred_var_epi) for epistemic-only CI\n", + " fmt=\"o\",\n", + " ecolor=\"lightgray\",\n", + " elinewidth=3,\n", + " capsize=0,\n", + " label=\"LOO prediction +/- 1.96*std_total\",\n", + " )\n", + " mn, mx = min(y_true.min(), y_pred.min()), max(y_true.max(), y_pred.max())\n", + " pad = 0.05 * (mx - mn)\n", + " diag = [mn - pad, mx + pad]\n", + " ax.plot(diag, diag, \"r--\", lw=2, label=\"Ideal\")\n", + " ax.set_xlabel(\"True value\")\n", + " ax.set_ylabel(\"GP LOO prediction\")\n", + " ax.set_title(f\"{display_name}\\nPredicted vs Actual\\nRMSE={loom.rmse:.4f}, MAE={loom.mae:.4f}\")\n", + " ax.legend(fontsize=8)\n", + "\n", + " # 4. Per-point epistemic vs aleatoric uncertainty bar plot\n", + " ax = axes[3, col]\n", + " x = np.arange(len(loo_results))\n", + " ax.bar(x, pred_var_epi, label=\"Epistemic Var\", alpha=0.6)\n", + " ax.bar(x, true_var_alea, bottom=pred_var_epi, label=\"Aleatoric Var\", alpha=0.6)\n", + " ax.set_xlabel(\"Left-out point index\")\n", + " ax.set_ylabel(\"Predictive Variance\")\n", + " ax.set_title(f\"{display_name}\\nEpistemic vs Aleatoric Uncertainty\")\n", + " ax.legend()\n", + "\n", + " fig.suptitle(f\"LOO Cross-Validation Results for {display_name}\", fontsize=16)\n", + " fig.tight_layout()\n", + " return fig\n" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + " Used same setup as doe.py to build experiment and extract training data.\n", + " import make_exp and add_sobol_points_to_exp from problem.py, and use extract_training_tensors to get (X, Y, Yvar) tensors.\n", + "\n", + "Later in another use case, we can load a SAVED experiment with data.\n", + "\n", + "### Interpretation of metrics:\n", + "#### Z_score:\n", + "\n", + "z-mean : should be close to 0, if the GP is unbiased in its predictions.\n", + "\n", + " - z-std : should be close to 1, indicates GP uncertarinty is well-calibrated.\n", + " - z-std > 1 : pred_var was lower, meaning GP is too certain than should be.\n", + " - z-std < 1 : pred_var was higher, meaning GP is too uncertain than should be.\n", + " - coverage_95 : should be close to 0.95, intervals contain true value 95% of the time." + ] + }, + { + "cell_type": "code", + "execution_count": 49, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "============================================================\n", + " GP LOO Cross-Validation: North Sea Crest Heights\n", + "============================================================\n", + "\n", + "STEP 2: Building experiment with 30 Sobol training points...\n", + " Experiment has 30 trials.\n", + "\n", + "STEP 3: Extracting training tensors from experiment...\n", + "Extracted raw data frame with 60 rows from Ax experiment.\n", + "Available metrics: \n", + "['loc', 'scale']\n", + "Length: 2, dtype: str\n", + "A row example: arm_name metric_name mean sem trial_index\n", + "0 0_0 loc 17.663189 0.251980 0\n", + "1 0_0 scale 1.075980 0.196031 0\n", + "2 1_0 loc 10.779497 0.197868 1\n", + "3 1_0 scale 0.839723 0.149940 1\n", + "4 2_0 loc 2.366150 0.033536 2\n", + "5 2_0 scale 0.142848 0.025974 2\n", + "6 3_0 loc 23.722449 0.403664 3\n", + "7 3_0 scale 1.708614 0.299496 3\n", + "8 4_0 loc 3.877968 0.047616 4\n", + "9 4_0 scale 0.202989 0.036090 4\n", + "10 5_0 loc 14.796933 0.231687 5\n", + "11 5_0 scale 0.986570 0.179717 5\n", + "12 6_0 loc 23.803095 0.381562 6\n", + "13 6_0 scale 1.622279 0.287940 6\n", + "14 7_0 loc 13.484618 0.191120 7\n", + "15 7_0 scale 0.818326 0.154005 7\n", + "16 8_0 loc 6.380567 0.123374 8\n", + "17 8_0 scale 0.526133 0.098433 8\n", + "18 9_0 loc 26.936055 0.473367 9\n", + "19 9_0 scale 2.000464 0.346752 9\n", + "20 10_0 loc 0.680377 0.009895 10\n", + "21 10_0 scale 0.041773 0.006910 10\n", + "22 11_0 loc 18.487342 0.281005 11\n", + "23 11_0 scale 1.193618 0.214549 11\n", + "24 12_0 loc 8.134274 0.124063 12\n", + "25 12_0 scale 0.523151 0.084533 12\n", + "26 13_0 loc 1.166147 0.013502 13\n", + "27 13_0 scale 0.056980 0.009725 13\n", + "28 14_0 loc 27.027586 0.350596 14\n", + "29 14_0 scale 1.482280 0.252829 14\n", + "30 15_0 loc 6.304950 0.094268 15\n", + "31 15_0 scale 0.400006 0.068510 15\n", + "32 16_0 loc 18.743906 0.395215 16\n", + "33 16_0 scale 1.685859 0.312435 16\n", + "34 17_0 loc 11.581708 0.160000 17\n", + "35 17_0 scale 0.673371 0.112171 17\n", + "36 18_0 loc 31.135395 0.467579 18\n", + "37 18_0 scale 1.981719 0.356995 18\n", + "38 19_0 loc 3.428204 0.047909 19\n", + "39 19_0 scale 0.203796 0.036721 19\n", + "40 20_0 loc 9.945529 0.175585 20\n", + "41 20_0 scale 0.742656 0.125283 20\n", + "42 21_0 loc 15.579711 0.274396 21\n", + "43 21_0 scale 1.165971 0.206285 21\n", + "44 22_0 loc 9.752168 0.166243 22\n", + "45 22_0 scale 0.703646 0.124619 22\n", + "46 23_0 loc 3.110305 0.066418 23\n", + "47 23_0 scale 0.279429 0.044745 23\n", + "48 24_0 loc 24.318752 0.436009 24\n", + "49 24_0 scale 1.837745 0.315444 24\n", + "50 25_0 loc 4.757033 0.081480 25\n", + "51 25_0 scale 0.346381 0.062590 25\n", + "52 26_0 loc 20.726003 0.449868 26\n", + "53 26_0 scale 1.900186 0.331845 26\n", + "54 27_0 loc 20.166019 0.360179 27\n", + "55 27_0 scale 1.528832 0.272458 27\n", + "56 28_0 loc 12.621135 0.253569 28\n", + "57 28_0 scale 1.070259 0.182407 28\n", + "58 29_0 loc 28.024439 0.619052 29\n", + "59 29_0 scale 2.649402 0.481835 29\n", + " X shape: torch.Size([30, 2]) (n_points, n_inputs)\n", + " Y shape: torch.Size([30, 2]) (n_points, n_outputs)\n", + " Yvar shape: torch.Size([30, 2]) (n_points, n_outputs)\n", + "\n", + " Y statistics (problem space):\n", + " loc: mean=13.9834, std=9.0347, noise_frac=3.19%\n", + " scale: mean=1.0130, std=0.6955, noise_frac=31.18%\n", + " bounds:\n", + " lower = [0.1, 1.0]\n", + " upper = [30.0, 30.0]\n", + "\n", + "STEP 4: Sanity checking GP fit on full data before LOO...\n", + " Mean abs error at training points (should be ~0):\n", + " loc: mean_abs_error=0.164206\n", + " scale: mean_abs_error=0.109643\n", + " Predictive variance vs Yvar at training points:\n", + " loc: mean_pred_var=0.122758, mean_Yvar=0.082926, ratio=148.03%\n", + " scale: mean_pred_var=0.055940, mean_Yvar=0.047045, ratio=118.91%\n", + "\n", + "STEP 5: Running Leave-One-Out cross validation on 30 points...\n", + "\n", + "Running LOO fold 1/30: leaving out point index 0...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.9834, 1.0130]), std = tensor([9.0347, 0.6955])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.8565, 1.0109]), std = tensor([9.1674, 0.7077])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 1/30: Left out point index 0, x=[[16.442171224858612, 13.204083887860179]]\n", + " True y=[17.663189308302176, 1.0759796653422808], sem²=[0.06349410275373502, 0.038428003078340336]\n", + " Predicted mean=[17.708135653985572, 1.1962009551384218]\n", + " Predicted epistemic var=[0.10645499325950425, 0.012709041416998135]\n", + " Predicted total var=[0.1900511596554515, 0.06005150326711686]\n", + "\n", + "Running LOO fold 2/30: leaving out point index 1...\n", + "Fold 2/30: Left out point index 1, x=[[11.173582748603074, 22.02443618979305]]\n", + " True y=[10.779497398513591, 0.8397225930615053], sem²=[0.03915168329741334, 0.02248206859805895]\n", + " Predicted mean=[10.661227314509288, 0.727436873352795]\n", + " Predicted epistemic var=[0.031579190407846625, 0.003525397184373058]\n", + " Predicted total var=[0.1160147505781498, 0.05141771884415666]\n", + "\n", + "Running LOO fold 3/30: leaving out point index 2...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.0939, 1.0190]), std = tensor([9.1740, 0.7071])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.3840, 1.0430]), std = tensor([8.9194, 0.6878])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 3/30: Left out point index 2, x=[[2.2488618210889397, 8.021291281096637]]\n", + " True y=[2.3661501302240957, 0.14284771832681287], sem²=[0.0011246309467408812, 0.0006746661015003941]\n", + " Predicted mean=[2.349252210413484, 0.13294997238357054]\n", + " Predicted epistemic var=[0.043653662362288514, 0.0011683211350497213]\n", + " Predicted total var=[0.12940050019985627, 0.049812622191266386]\n", + "\n", + "Running LOO fold 4/30: leaving out point index 3...\n", + "Fold 4/30: Left out point index 3, x=[[23.37228300916031, 28.170240418985486]]\n", + " True y=[23.72244894701183, 1.708613957450132], sem²=[0.16294428375062836, 0.08969789685294823]\n", + " Predicted mean=[23.271992087616525, 1.7848580255331346]\n", + " Predicted epistemic var=[0.2201080787527161, 0.024652709385335925]\n", + " Predicted total var=[0.30027492856256355, 0.07022724386391646]\n", + "\n", + "Running LOO fold 5/30: leaving out point index 4...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.6476, 0.9890]), std = tensor([9.0021, 0.6951])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.3318, 1.0410]), std = tensor([8.9871, 0.6905])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 5/30: Left out point index 4, x=[[4.164683258719742, 25.320823485963047]]\n", + " True y=[3.8779680552224858, 0.20298929296725668], sem²=[0.0022672400679046973, 0.0013025230614270454]\n", + " Predicted mean=[3.7472688359830393, 0.23276613334615126]\n", + " Predicted epistemic var=[0.011248096943759833, 0.0010672037490375708]\n", + " Predicted total var=[0.09695553446680467, 0.04968985456525676]\n", + "\n", + "Running LOO fold 6/30: leaving out point index 5...\n", + "Fold 6/30: Left out point index 5, x=[[12.994798111543059, 10.382323726080358]]\n", + " True y=[14.796932822783514, 0.9865704569530516], sem²=[0.05367891048091574, 0.032298193471728874]\n", + " Predicted mean=[13.985604291227112, 0.9164929670783958]\n", + " Predicted epistemic var=[0.12779593148118806, 0.013760121155472316]\n", + " Predicted total var=[0.2117305527830946, 0.06131395575064662]\n", + "\n", + "Running LOO fold 7/30: leaving out point index 6...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.9553, 1.0139]), std = tensor([9.1933, 0.7078])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.6448, 0.9920]), std = tensor([8.9988, 0.6981])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 7/30: Left out point index 6, x=[[21.971921502798796, 15.606687484309077]]\n", + " True y=[23.803094852178575, 1.622279416306266], sem²=[0.14558991804203614, 0.08290928803295416]\n", + " Predicted mean=[22.66179225032921, 1.5149527714002566]\n", + " Predicted epistemic var=[0.15927951075940427, 0.018145543317885426]\n", + " Predicted total var=[0.24004478697299633, 0.06395416775577609]\n", + "\n", + "Running LOO fold 8/30: leaving out point index 7...\n", + "Fold 8/30: Left out point index 7, x=[[14.29993766117841, 29.01312247198075]]\n", + " True y=[13.484617625637396, 0.8183261352983292], sem²=[0.03652702160939107, 0.023717526342367474]\n", + " Predicted mean=[13.836762472516348, 1.086243828764492]\n", + " Predicted epistemic var=[0.16000220033954804, 0.015067417310300701]\n", + " Predicted total var=[0.24452826608530026, 0.06291713697890126]\n", + "\n", + "Running LOO fold 9/30: leaving out point index 8...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.0006, 1.0197]), std = tensor([9.1941, 0.7069])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.2455, 1.0298]), std = tensor([9.0778, 0.7016])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 9/30: Left out point index 8, x=[[6.31648598127067, 14.046260482631624]]\n", + " True y=[6.380567295691077, 0.5261331448351754], sem²=[0.015221256353744965, 0.009689083431754432]\n", + " Predicted mean=[6.13564301527755, 0.45909531820478744]\n", + " Predicted epistemic var=[0.02213834422945382, 0.0019299492277221608]\n", + " Predicted total var=[0.10739909153574556, 0.05026340830703351]\n", + "\n", + "Running LOO fold 10/30: leaving out point index 9...\n", + "Fold 10/30: Left out point index 9, x=[[26.27590978220105, 19.242442643269897]]\n", + " True y=[26.936055494497538, 2.000464011856637], sem²=[0.22407634318243883, 0.12023697596556285]\n", + " Predicted mean=[27.408941188292143, 1.7530766338860257]\n", + " Predicted epistemic var=[0.09151765501761844, 0.02250841620441002]\n", + " Predicted total var=[0.16957650277809314, 0.06702987898945212]\n", + "\n", + "Running LOO fold 11/30: leaving out point index 10...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.5367, 0.9790]), std = tensor([8.8512, 0.6819])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.4421, 1.0465]), std = tensor([8.8319, 0.6828])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 11/30: Left out point index 10, x=[[0.7503059620968997, 18.26130876597017]]\n", + " True y=[0.6803771012380035, 0.041773016145446075], sem²=[9.790599871215454e-05, 4.7742847693033794e-05]\n", + " Predicted mean=[0.618931058604538, 0.057535317901073135]\n", + " Predicted epistemic var=[0.22373261253780186, 0.007760000332989403]\n", + " Predicted total var=[0.3095148546839223, 0.056425919432440796]\n", + "\n", + "Running LOO fold 12/30: leaving out point index 11...\n", + "Fold 12/30: Left out point index 11, x=[[18.33480247752741, 24.35032025910914]]\n", + " True y=[18.487341953646883, 1.1936183821226922], sem²=[0.07896363173959582, 0.04603133615761128]\n", + " Predicted mean=[17.89113407429522, 1.3951633434377573]\n", + " Predicted epistemic var=[0.07303467525098029, 0.010592607254579]\n", + " Predicted total var=[0.15609740961293234, 0.057672885205412534]\n", + "\n", + "Running LOO fold 13/30: leaving out point index 12...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.8281, 1.0068]), std = tensor([9.1538, 0.7070])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.1851, 1.0299]), std = tensor([9.1256, 0.7016])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 13/30: Left out point index 12, x=[[8.274001326691359, 23.0941325686872]]\n", + " True y=[8.134273729749243, 0.5231505957201273], sem²=[0.015391679905549054, 0.007145902020077193]\n", + " Predicted mean=[7.724979163380697, 0.5043781877863834]\n", + " Predicted epistemic var=[0.015635036633995014, 0.002574131610923125]\n", + " Predicted total var=[0.10088990726608661, 0.050995286600981966]\n", + "\n", + "Running LOO fold 14/30: leaving out point index 13...\n", + "Fold 14/30: Left out point index 13, x=[[1.1905946464277803, 9.090213060379028]]\n", + " True y=[1.1661474175591287, 0.05697999474835876], sem²=[0.00018231220471421277, 9.456671124936534e-05]\n", + " Predicted mean=[1.287097733269496, 0.0870917699839533]\n", + " Predicted epistemic var=[0.04382177675858401, 0.0014417771620885977]\n", + " Predicted total var=[0.1296011083458768, 0.050106081645555293]\n", + "\n", + "Running LOO fold 15/30: leaving out point index 14...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.4254, 1.0460]), std = tensor([8.8584, 0.6836])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.5336, 0.9969]), std = tensor([8.8462, 0.7021])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 15/30: Left out point index 14, x=[[25.84756417470053, 17.91125872079283]]\n", + " True y=[27.027586049707175, 1.482280183126373], sem²=[0.12291764013101472, 0.06392251270687722]\n", + " Predicted mean=[26.775530609926527, 1.8738579179302017]\n", + " Predicted epistemic var=[0.1607278877491254, 0.028697023239885877]\n", + " Predicted total var=[0.24227496664930437, 0.07516036406833093]\n", + "\n", + "Running LOO fold 16/30: leaving out point index 15...\n", + "Fold 16/30: Left out point index 15, x=[[6.8736472832038995, 20.49868467450142]]\n", + " True y=[6.304950290116001, 0.4000058962000265], sem²=[0.008886535923932268, 0.004693684533390071]\n", + " Predicted mean=[6.6487933877704535, 0.44850017017658417]\n", + " Predicted epistemic var=[0.014362154400949123, 0.0021471212309553955]\n", + " Predicted total var=[0.09984134034275165, 0.05065283544469311]\n", + "\n", + "Running LOO fold 17/30: leaving out point index 16...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.2482, 1.0342]), std = tensor([9.0754, 0.6980])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.8192, 0.9898]), std = tensor([9.1490, 0.6959])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 17/30: Left out point index 16, x=[[18.83865722194314, 29.363226857967675]]\n", + " True y=[18.74390562231763, 1.685859351765245], sem²=[0.15619514716885277, 0.09761535039054148]\n", + " Predicted mean=[18.489908825495032, 1.347965250619219]\n", + " Predicted epistemic var=[0.16064216446051205, 0.017601427937575798]\n", + " Predicted total var=[0.24104174311800697, 0.06290294677692898]\n", + "\n", + "Running LOO fold 18/30: leaving out point index 17...\n", + "Fold 18/30: Left out point index 17, x=[[11.619274637289344, 17.068307526409626]]\n", + " True y=[11.581707569723642, 0.6733711285653953], sem²=[0.025599956015515263, 0.012582223599092897]\n", + " Predicted mean=[11.357644158012311, 0.8659270092760462]\n", + " Predicted epistemic var=[0.04661249271283907, 0.004506599546588164]\n", + " Predicted total var=[0.13151535382389737, 0.052740295171853364]\n", + "\n", + "Running LOO fold 19/30: leaving out point index 18...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.0662, 1.0247]), std = tensor([9.1830, 0.7048])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.3919, 0.9796]), std = tensor([8.5833, 0.6829])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 19/30: Left out point index 18, x=[[29.43744654878974, 25.87608463037759]]\n", + " True y=[31.1353951876523, 1.9817192939071568], sem²=[0.21863036265454888, 0.12744524613577565]\n", + " Predicted mean=[29.092287834551477, 1.9761662184681925]\n", + " Predicted epistemic var=[0.6778856535880848, 0.05777081148973573]\n", + " Predicted total var=[0.7561322937805558, 0.10204371323442565]\n", + "\n", + "Running LOO fold 20/30: leaving out point index 19...\n", + "Fold 20/30: Left out point index 19, x=[[3.6781598697416484, 26.708564091473818]]\n", + " True y=[3.4282043658790085, 0.2037959862070608], sem²=[0.0022952608278597212, 0.0013484534489774061]\n", + " Predicted mean=[3.6074218127451694, 0.1731195860462713]\n", + " Predicted epistemic var=[0.020462691274758527, 0.0023351359052977694]\n", + " Predicted total var=[0.10616916256470148, 0.05095620291504971]\n", + "\n", + "Running LOO fold 21/30: leaving out point index 20...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.3474, 1.0409]), std = tensor([8.9680, 0.6905])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.1226, 1.0224]), std = tensor([9.1618, 0.7059])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 21/30: Left out point index 20, x=[[9.681309627089648, 12.61961493641138]]\n", + " True y=[9.945528549097316, 0.7426562805067928], sem²=[0.030830190310578338, 0.01569593563337351]\n", + " Predicted mean=[10.42057220974159, 0.7059561882761414]\n", + " Predicted epistemic var=[0.05586358915708445, 0.005306221462274552]\n", + " Predicted total var=[0.14058609736141642, 0.05343254770704732]\n", + "\n", + "Running LOO fold 22/30: leaving out point index 21...\n", + "Fold 22/30: Left out point index 21, x=[[15.887822312582284, 21.469118759967387]]\n", + " True y=[15.57971076709207, 1.1659706213642982], sem²=[0.07529295538822185, 0.042553474540788894]\n", + " Predicted mean=[15.558583498890288, 1.1349761951520918]\n", + " Predicted epistemic var=[0.0655859017851057, 0.007344075218215562]\n", + " Predicted total var=[0.14877521119365686, 0.054544279431698144]\n", + "\n", + "Running LOO fold 23/30: leaving out point index 22...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.9283, 1.0078]), std = tensor([9.1895, 0.7072])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.1293, 1.0237]), std = tensor([9.1586, 0.7053])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 23/30: Left out point index 22, x=[[10.182581103220581, 27.72896001394838]]\n", + " True y=[9.752167718782093, 0.7036464092266824], sem²=[0.027636865347741152, 0.015529837148773715]\n", + " Predicted mean=[10.004512974050538, 0.6054057224132537]\n", + " Predicted epistemic var=[0.10725582684735713, 0.007362906882338316]\n", + " Predicted total var=[0.19208844970558003, 0.0554949606610628]\n", + "\n", + "Running LOO fold 24/30: leaving out point index 23...\n", + "Fold 24/30: Left out point index 23, x=[[3.2503646770492196, 13.638254883699119]]\n", + " True y=[3.1103053049085645, 0.2794289467405284], sem²=[0.004411389497209105, 0.0020021543314199743]\n", + " Predicted mean=[3.028393774233521, 0.20666679455773818]\n", + " Predicted epistemic var=[0.02683588064351028, 0.0013809870627631282]\n", + " Predicted total var=[0.11246938197933774, 0.04997951266277567]\n", + "\n", + "Running LOO fold 25/30: leaving out point index 24...\n", + "Fold 25/30: Left out point index 24, x=[[24.37822638656944, 22.489556113258004]]\n", + " True y=[24.31875249682041, 1.8377450925898977], sem²=[0.19010416675712308, 0.09950497380802684]\n", + " Predicted mean=[25.21134086911721, 1.7697166616123763]\n", + " Predicted epistemic var=[0.09755902631184199, 0.014454191107791226]\n", + " Predicted total var=[0.17678932843181033, 0.05969055051895525]\n", + "\n", + "Running LOO fold 26/30: leaving out point index 25...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.3583, 1.0383]), std = tensor([8.9540, 0.6937])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.6270, 0.9846]), std = tensor([8.9775, 0.6899])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.3015, 1.0360]), std = tensor([9.0220, 0.6962])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 26/30: Left out point index 25, x=[[5.042766495887189, 16.049654993228614]]\n", + " True y=[4.757033214712732, 0.3463812244347482], sem²=[0.006639071830514091, 0.003917470028461905]\n", + " Predicted mean=[4.79048349203244, 0.3646730116465081]\n", + " Predicted epistemic var=[0.011424602854731347, 0.0016026630852544743]\n", + " Predicted total var=[0.09698128755837587, 0.050135143316403494]\n", + "\n", + "Running LOO fold 27/30: leaving out point index 26...\n", + "Fold 27/30: Left out point index 26, x=[[21.10855199014768, 24.857445871457458]]\n", + " True y=[20.726002895825605, 1.9001864006869897], sem²=[0.20238163628485648, 0.11012114544297735]\n", + " Predicted mean=[21.381750374298484, 1.5561094309524375]\n", + " Predicted epistemic var=[0.08115439176017958, 0.01180032924866381]\n", + " Predicted total var=[0.15996133286195025, 0.05667061377586402]\n", + "\n", + "Running LOO fold 28/30: leaving out point index 27...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.7509, 0.9824]), std = tensor([9.1028, 0.6870])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.7702, 0.9952]), std = tensor([9.1175, 0.7009])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 28/30: Left out point index 27, x=[[19.27474700929597, 14.481136666610837]]\n", + " True y=[20.16601894743115, 1.5288317336323933], sem²=[0.12972890967030512, 0.07423358581976085]\n", + " Predicted mean=[20.805399368741938, 1.327986068051318]\n", + " Predicted epistemic var=[0.09944405083225405, 0.014649369118695965]\n", + " Predicted total var=[0.18075625836900924, 0.06075715570186917]\n", + "\n", + "Running LOO fold 29/30: leaving out point index 28...\n", + "Fold 29/30: Left out point index 28, x=[[13.304936639312654, 19.706608469597995]]\n", + " True y=[12.621135414756644, 1.07025900321407], sem²=[0.06429701103080432, 0.03327224387361257]\n", + " Predicted mean=[13.090611770583566, 0.9064401974722095]\n", + " Predicted epistemic var=[0.03075766421302717, 0.004905245350140919]\n", + " Predicted total var=[0.1143261441166617, 0.05242549200042268]\n", + "\n", + "Running LOO fold 30/30: leaving out point index 29...\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([14.0304, 1.0111]), std = tensor([9.1909, 0.7078])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n", + "c:\\Users\\ANUVATS\\Documents\\Work\\Projects\\axtreme\\axtreme\\.venv\\Lib\\site-packages\\botorch\\models\\utils\\assorted.py:202: InputDataWarning: Input data is not standardized (mean = tensor([13.4992, 0.9566]), std = tensor([8.7896, 0.6341])). Please consider scaling the input to zero mean and unit variance.\n", + " warnings.warn(msg, InputDataWarning)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fold 30/30: Left out point index 29, x=[[27.25596907204017, 28.571136770769954]]\n", + " True y=[28.02443861951843, 2.6494023466665095], sem²=[0.3832249090636081, 0.23216533261665973]\n", + " Predicted mean=[28.31173512792255, 1.8216998230502286]\n", + " Predicted epistemic var=[0.2692203907868418, 0.0379545602730314]\n", + " Predicted total var=[0.34179135696520724, 0.07861642455286325]\n" + ] + } + ], + "source": [ + "print(\"=\" * 60)\n", + "print(\" GP LOO Cross-Validation: North Sea Crest Heights\")\n", + "print(\"=\" * 60)\n", + "\n", + "from axtreme.experiment import add_sobol_points_to_experiment # type: ignore[import-not-found] # noqa: E402\n", + "\n", + "# N_TRAINING_POINTS: total points in GP training set.\n", + "# N_LOO_POINTS: folds to evaluate; set equal to N_TRAINING_POINTS to run all.\n", + "\n", + "\n", + "def print_summary(metrics_list: list[LOOMetrics]) -> None:\n", + " \"\"\"Print a nicely formatted summary of LOO metrics.\"\"\"\n", + " print(\"\\n\" + \"=\" * 65)\n", + " print(\"\\nLOO Metrics Summary:\")\n", + " print(\"\\n\" + \"=\" * 65)\n", + " for m in metrics_list:\n", + " print(f\"Metric: {m.metric_name}\")\n", + " print(f\" MAE: {m.mae:.4f}\")\n", + " print(f\" RMSE: {m.rmse:.4f}\")\n", + " print(f\" Z-score mean: {m.z_mean:.4f} (should be close to 0)\")\n", + " print(f\" Z-score std: {m.z_std:.4f} (should be close to 1)\")\n", + " print(f\" 95% CI coverage: {m.ci_95_coverage:.2%} (should be close to 95%)\")\n", + " print(f\" Fraction aleatoric variance: {m.frac_aleatoric:.2%}\")\n", + "\n", + " for m in metrics_list:\n", + " print(f\"\\n For metric {m.metric_name}, the GP's predictive uncertainty is:\")\n", + " if m.z_std > 1:\n", + " print(f\" too certain (z-std={m.z_std:.2f} > 1), predictive variance lower than actual error.\")\n", + " elif m.z_std < 1:\n", + " print(f\" too uncertain (z-std={m.z_std:.2f} < 1), predictive variance higher than actual error.\")\n", + " else:\n", + " print(f\" well-calibrated (z-std={m.z_std:.2f} ~ 1), predictive variance matches actual error.\")\n", + "\n", + " # Decompose total GP uncertainty into aleatoric and epistemic parts.\n", + " # High aleatoric -> need more simulations per point.\n", + " # High epistemic -> benefit from adding more training points nearby.\n", + " for m in metrics_list:\n", + " mean_frac = float(np.mean(m.frac_aleatoric))\n", + " print(f\" {m.metric_name}: {mean_frac:.1%} aleatoric, {1 - mean_frac:.1%} epistemic\")\n", + "\n", + "\n", + "N_TRAINING_POINTS = 30 # How many training points to put in the GP\n", + "N_LOO_POINTS = 30 # How many folds to evaluate (set == N_TRAINING to run all)\n", + "SEED_TRAINING = 8 # Sobol seed (matches doe.py)\n", + "METRIC_NAMES = [\"loc\", \"scale\"] # output metrics (Gumbel distribution params)\n", + "PARAMETER_NAMES = [\"Hs\", \"Tp\"] # input parameters (search space dimensions)\n", + "METRIC_DISPLAY_NAMES = [\"Gumbel loc\", \"Gumbel scale\"]\n", + "\n", + "print(f\"\\nSTEP 2: Building experiment with {N_TRAINING_POINTS} Sobol training points...\")\n", + "exp = make_exp()\n", + "add_sobol_points_to_experiment(exp, n_iter=N_TRAINING_POINTS, seed=SEED_TRAINING)\n", + "print(f\" Experiment has {len(exp.trials)} trials.\")\n", + "\n", + "print(\"\\nSTEP 3: Extracting training tensors from experiment...\")\n", + "X, Y, Yvar = extract_training_tensors(exp, METRIC_NAMES, PARAMETER_NAMES)\n", + "\n", + "print(f\" X shape: {X.shape} (n_points, n_inputs)\")\n", + "print(f\" Y shape: {Y.shape} (n_points, n_outputs)\")\n", + "print(f\" Yvar shape: {Yvar.shape} (n_points, n_outputs)\")\n", + "print(\"\\n Y statistics (problem space):\")\n", + "for i, mn in enumerate(METRIC_NAMES):\n", + " print(\n", + " f\" {mn}: mean={Y[:, i].mean():.4f}, std={Y[:, i].std():.4f}, \"\n", + " f\"noise_frac={Yvar[:, i].mean().sqrt() / Y[:, i].std():.2%}\"\n", + " )\n", + "\n", + "bounds = make_bounds_tensor(SEARCH_SPACE, PARAMETER_NAMES)\n", + "print(f\" bounds:\\n lower = {bounds[0].tolist()}\\n upper = {bounds[1].tolist()}\")\n", + "\n", + "# Sanity check: fit GP on full data before LOO to confirm data extraction is correct.\n", + "# Expected: posterior mean at training points ~ train_Y; epistemic variance ~ 0.\n", + "print(\"\\nSTEP 4: Sanity checking GP fit on full data before LOO...\")\n", + "full_gp = build_gp(X, Y, Yvar, bounds, fit=True)\n", + "\n", + "with torch.no_grad():\n", + " posterior_full = full_gp.posterior(X, observation_noise=True)\n", + " pred_mean_full = posterior_full.mean # (n, m)\n", + " pred_var_full = posterior_full.variance # (n, m)\n", + "\n", + "mean_error = (pred_mean_full - Y).abs().mean(dim=0)\n", + "print(\" Mean abs error at training points (should be ~0):\")\n", + "for i, mn in enumerate(METRIC_NAMES):\n", + " print(f\" {mn}: mean_abs_error={mean_error[i]:.6f}\")\n", + "print(\" Predictive variance vs Yvar at training points:\")\n", + "for i, mn in enumerate(METRIC_NAMES):\n", + " print(\n", + " f\" {mn}: mean_pred_var={pred_var_full[:, i].mean():.6f}, \"\n", + " f\"mean_Yvar={Yvar[:, i].mean():.6f}, \"\n", + " f\"ratio={pred_var_full[:, i].mean() / Yvar[:, i].mean():.2%}\"\n", + " )\n", + "\n", + "# Run LOO validation loop.\n", + "loo_indices = list(range(min(N_LOO_POINTS, X.shape[0])))\n", + "print(f\"\\nSTEP 5: Running Leave-One-Out cross validation on {len(loo_indices)} points...\")\n", + "loo_results = run_loo(X, Y, Yvar, bounds, indices=loo_indices, verbose=True)\n" + ] + }, + { + "cell_type": "code", + "execution_count": 50, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "\n", + "STEP 6: Computing LOO metrics...\n", + "\n", + "=================================================================\n", + "\n", + "LOO Metrics Summary:\n", + "\n", + "=================================================================\n", + "Metric: loc\n", + " MAE: 0.4025\n", + " RMSE: 0.5740\n", + " Z-score mean: -0.0215 (should be close to 0)\n", + " Z-score std: 1.1182 (should be close to 1)\n", + " 95% CI coverage: 90.00% (should be close to 95%)\n", + " Fraction aleatoric variance: 54.31%\n", + "Metric: scale\n", + " MAE: 0.1414\n", + " RMSE: 0.2184\n", + " Z-score mean: -0.1914 (should be close to 0)\n", + " Z-score std: 0.8090 (should be close to 1)\n", + " 95% CI coverage: 96.67% (should be close to 95%)\n", + " Fraction aleatoric variance: 82.35%\n", + "\n", + " For metric loc, the GP's predictive uncertainty is:\n", + " too certain (z-std=1.12 > 1), predictive variance lower than actual error.\n", + "\n", + " For metric scale, the GP's predictive uncertainty is:\n", + " too uncertain (z-std=0.81 < 1), predictive variance higher than actual error.\n", + " loc: 54.3% aleatoric, 45.7% epistemic\n", + " scale: 82.4% aleatoric, 17.6% epistemic\n", + "\n", + "STEP 7: Plotting LOO results...\n" + ] + }, + { + "data": { + "image/png": "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", + "text/plain": [ + "
" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "\n", + "LOO results plot saved to results\\GP_eval_results\\loo_results_30_points.png\n" + ] + } + ], + "source": [ + "# Key metrics per output metric (loc and scale):\n", + "# 1. RMSE of predicted mean vs true value at left-out point.\n", + "# 2. Z-score = (y_pred_mean - y_left_out) / sqrt(pred_var_total)\n", + "# - should be ~ N(0, 1) if uncertainties are well calibrated.\n", + "# - Var[z] > 1: predictive variance too small, GP is overconfident.\n", + "# - Var[z] < 1: predictive variance too large, GP is underconfident.\n", + "# 3. 95% CI coverage: fraction of held-out points where true value falls inside\n", + "# [mu ± 1.96*sigma]. Should be close to 0.95.\n", + "# 4. Epistemic vs aleatoric split:\n", + "# - High epistemic at a point -> more training data nearby would help.\n", + "# - High aleatoric -> need more simulations per point to reduce label noise.\n", + "\n", + "print(\"\\nSTEP 6: Computing LOO metrics...\")\n", + "loo_metrics_list = [compute_metrics(loo_results, metric_idx=i, metric_name=mn) for i, mn in enumerate(METRIC_NAMES)]\n", + "\n", + "print_summary(loo_metrics_list)\n", + "\n", + "print(\"\\nSTEP 7: Plotting LOO results...\")\n", + "fig = plot_loo_results(loo_results, loo_metrics_list, METRIC_DISPLAY_NAMES)\n", + "plt.show()\n", + "\n", + "# save plots\n", + "results_dir = Path(\"results/GP_eval_results\")\n", + "results_dir.mkdir(parents=True, exist_ok=True)\n", + "fig.savefig(results_dir / f\"loo_results_{N_TRAINING_POINTS}_points.png\")\n", + "print(f\"\\nLOO results plot saved to {results_dir / f'loo_results_{N_TRAINING_POINTS}_points.png'}\")\n" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "axtreme (3.12.12)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.12.12" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +}