Add EmpiricalCovariance algorithm

configs/common/covariance.json

@@ -0,0 +1,15 @@
+{
+    "INCLUDE": ["../common/sklearn.json"],
+    "PARAMETERS_SETS": {
+        "covariance parameters": {
+            "algorithm": {
+                "estimator": "EmpiricalCovariance",
+                "estimator_params": {
+                    "store_precision": [true, false],
+                    "assume_centered": false
+                },
+                "estimator_methods": { "training": "fit" }
+            }
+        }
+    }
+}

configs/common/sklearn.json

@@ -18,6 +18,24 @@
                 "data": { "format": "dpnp", "order": "C" }
             }
         ],
+        "sklearn-ex[cpu] preview implementations": {
+            "algorithm": [
+                { "library": "sklearn", "device": "cpu" },
+                { "library": "sklearnex.preview", "device": "cpu" }
+            ]
+        },
+        "sklearn-ex[cpu,gpu] preview implementations": [
+            { "algorithm": { "library": "sklearn", "device": "cpu" } },
+            { "algorithm": { "library": "sklearnex.preview", "device": "cpu" } },
+            {
+                "algorithm": {
+                    "library": "sklearnex.preview",
+                    "device": "gpu",
+                    "sklearn_context": { "array_api_dispatch": true }
+                },
+                "data": { "format": "dpnp", "order": "C" }
+            }
+        ],
         "sklearnex spmd implementation": {
             "algorithm": {
                 "library": "sklearnex.spmd",

configs/experiments/covariance.json

@@ -0,0 +1,29 @@
+{
+    "INCLUDE": ["../common/sklearn.json", "../common/covariance.json"],
+    "PARAMETERS_SETS": {
+        "covariance datasets": [
+            {
+                "data": {
+                    "source": "make_blobs",
+                    "generation_kwargs": [
+                        { "n_samples": 10000000, "n_features": 50, "centers": 1 },
+                        { "n_samples": 2000000, "n_features": 500, "centers": 1 },
+                        { "n_samples": 200000, "n_features": 2000, "centers": 1 }
+                    ],
+                    "split_kwargs": { "ignore": true }
+                }
+            },
+            {
+                "data": {
+                    "dataset": ["susy", "higgs"],
+                    "split_kwargs": { "ignore": true }
+                }
+            }
+        ]
+    },
+    "TEMPLATES": {
+        "sklearn covariance": {
+            "SETS": ["sklearn-ex[cpu,gpu] preview implementations", "covariance parameters", "covariance datasets"]
+        }
+    }
+}

configs/weekly/covariance.json

@@ -0,0 +1,33 @@
+{
+    "INCLUDE": ["../common/sklearn.json", "../common/covariance.json"],
+    "PARAMETERS_SETS": {
+        "high-load covariance datasets": [
+            {
+                "data": {
+                    "source": "make_blobs",
+                    "generation_kwargs": [
+                        { "n_samples": 50000000, "n_features": 50, "centers": 1 },
+                        { "n_samples": 10000000, "n_features": 500, "centers": 1 },
+                        { "n_samples": 1000000, "n_features": 5000, "centers": 1 }
+                    ],
+                    "split_kwargs": { "ignore": true }
+                }
+            },
+            {
+                "data": {
+                    "dataset": ["airline_depdelay", "epsilon"],
+                    "split_kwargs": { "ignore": true }
+                }
+            }
+        ]
+    },
+    "TEMPLATES": {
+        "sklearn covariance": {
+            "SETS": [
+                "sklearn-ex[cpu,gpu] preview implementations",
+                "covariance parameters",
+                "high-load covariance datasets"
+            ]
+        }
+    }
+}

sklbench/benchmarks/sklearn_estimator.py

@@ -190,6 +190,15 @@ def get_subset_metrics_of_estimator(
                         )
                     }
                 )
+    elif task == "utility":
+        # SPMD estimators are skipped: their score triggers an additional collective fit
+        if (
+            "Covariance" in str(estimator_instance)
+            and stage == "training"
+            and hasattr(estimator_instance, "score")
+            and "spmd" not in type(estimator_instance).__module__
+        ):
+            metrics.update({"average log-likelihood": float(estimator_instance.score(x))})
     elif task == "clustering":
         if hasattr(estimator_instance, "inertia_"):
             # compute inertia manually using distances to cluster centers

sklbench/report/implementation.py

@@ -74,7 +74,7 @@
     "indifferent": [
         # SVM
         "support vectors",
-        # PCA
+        # PCA, EmpiricalCovariance
         "average log-likelihood",
         "1st component variance ratio",
         # DBSCAN