Sarah Segel: Handle configspace as dictionary in mlp example (#1057)

development/.buildinfo

@@ -1,4 +1,4 @@
 # Sphinx build info version 1
 # This file hashes the configuration used when building these files. When it is not found, a full rebuild will be done.
-config: 7998ea861451546be11046d510c9eb9f
+config: 9e450612d63cf85af7583a34de820462
 tags: 645f666f9bcd5a90fca523b33c5a78b7

development/_downloads/0422a2b52bca2c17c0ab975a6ae7e8db/2_parego.py

@@ -66,9 +66,9 @@ def configspace(self) -> ConfigurationSpace:
         return cs
 
     def train(self, config: Configuration, seed: int = 0, budget: int = 10) -> dict[str, float]:
-        lr = config["learning_rate"] if config["learning_rate"] else "constant"
-        lr_init = config["learning_rate_init"] if config["learning_rate_init"] else 0.001
-        batch_size = config["batch_size"] if config["batch_size"] else 200
+        lr = config.get("learning_rate", "constant")
+        lr_init = config.get("learning_rate_init", 0.001)
+        batch_size = config.get("batch_size", 200)
 
         start_time = time.time()
 

development/_downloads/88e1c1ee9fc33236ebfa159898782a73/2_sgd_datasets.py

@@ -89,7 +89,7 @@ def train(self, config: Configuration, instance: str, seed: int = 0) -> float:
 
             # SGD classifier using given configuration
             clf = SGDClassifier(
-                loss="log",
+                loss="log_loss",
                 penalty="elasticnet",
                 alpha=config["alpha"],
                 l1_ratio=config["l1_ratio"],

development/_downloads/8a5c5b9a1383ea80dbf4608b637a4168/2_sgd_datasets.ipynb

@@ -15,7 +15,7 @@
       },
       "outputs": [],
       "source": [
-        "from __future__ import annotations\n\nimport itertools\nimport warnings\n\nimport numpy as np\nfrom ConfigSpace import Categorical, Configuration, ConfigurationSpace, Float\nfrom sklearn import datasets\nfrom sklearn.linear_model import SGDClassifier\nfrom sklearn.model_selection import StratifiedKFold, cross_val_score\n\nfrom smac import MultiFidelityFacade as MFFacade\nfrom smac import Scenario\n\n__copyright__ = \"Copyright 2021, AutoML.org Freiburg-Hannover\"\n__license__ = \"3-clause BSD\"\n\n\nclass DigitsDataset:\n    def __init__(self) -> None:\n        self._data = datasets.load_digits()\n\n    def get_instances(self) -> list[str]:\n        \"\"\"Create instances from the dataset which include two classes only.\"\"\"\n        return [f\"{classA}-{classB}\" for classA, classB in itertools.combinations(self._data.target_names, 2)]\n\n    def get_instance_features(self) -> dict[str, list[int | float]]:\n        \"\"\"Returns the mean and variance of all instances as features.\"\"\"\n        features = {}\n        for instance in self.get_instances():\n            data, _ = self.get_instance_data(instance)\n            features[instance] = [np.mean(data), np.var(data)]\n\n        return features\n\n    def get_instance_data(self, instance: str) -> tuple[np.ndarray, np.ndarray]:\n        \"\"\"Retrieve data from the passed instance.\"\"\"\n        # We split the dataset into two classes\n        classA, classB = instance.split(\"-\")\n        indices = np.where(np.logical_or(int(classA) == self._data.target, int(classB) == self._data.target))\n\n        data = self._data.data[indices]\n        target = self._data.target[indices]\n\n        return data, target\n\n\nclass SGD:\n    def __init__(self, dataset: DigitsDataset) -> None:\n        self.dataset = dataset\n\n    @property\n    def configspace(self) -> ConfigurationSpace:\n        \"\"\"Build the configuration space which defines all parameters and their ranges for the SGD classifier.\"\"\"\n        cs = ConfigurationSpace()\n\n        # We define a few possible parameters for the SGD classifier\n        alpha = Float(\"alpha\", (0, 1), default=1.0)\n        l1_ratio = Float(\"l1_ratio\", (0, 1), default=0.5)\n        learning_rate = Categorical(\"learning_rate\", [\"constant\", \"invscaling\", \"adaptive\"], default=\"constant\")\n        eta0 = Float(\"eta0\", (0.00001, 1), default=0.1, log=True)\n        # Add the parameters to configuration space\n        cs.add_hyperparameters([alpha, l1_ratio, learning_rate, eta0])\n\n        return cs\n\n    def train(self, config: Configuration, instance: str, seed: int = 0) -> float:\n        \"\"\"Creates a SGD classifier based on a configuration and evaluates it on the\n        digits dataset using cross-validation.\"\"\"\n\n        with warnings.catch_warnings():\n            warnings.filterwarnings(\"ignore\")\n\n            # SGD classifier using given configuration\n            clf = SGDClassifier(\n                loss=\"log\",\n                penalty=\"elasticnet\",\n                alpha=config[\"alpha\"],\n                l1_ratio=config[\"l1_ratio\"],\n                learning_rate=config[\"learning_rate\"],\n                eta0=config[\"eta0\"],\n                max_iter=30,\n                early_stopping=True,\n                random_state=seed,\n            )\n\n            # get instance\n            data, target = self.dataset.get_instance_data(instance)\n\n            cv = StratifiedKFold(n_splits=4, random_state=seed, shuffle=True)  # to make CV splits consistent\n            scores = cross_val_score(clf, data, target, cv=cv)\n\n        return 1 - np.mean(scores)\n\n\nif __name__ == \"__main__\":\n    dataset = DigitsDataset()\n    model = SGD(dataset)\n\n    scenario = Scenario(\n        model.configspace,\n        walltime_limit=30,  # We want to optimize for 30 seconds\n        n_trials=5000,  # We want to try max 5000 different trials\n        min_budget=1,  # Use min one instance\n        max_budget=45,  # Use max 45 instances (if we have a lot of instances we could constraint it here)\n        instances=dataset.get_instances(),\n        instance_features=dataset.get_instance_features(),\n    )\n\n    # Create our SMAC object and pass the scenario and the train method\n    smac = MFFacade(\n        scenario,\n        model.train,\n        overwrite=True,\n    )\n\n    # Now we start the optimization process\n    incumbent = smac.optimize()\n\n    default_cost = smac.validate(model.configspace.get_default_configuration())\n    print(f\"Default cost: {default_cost}\")\n\n    incumbent_cost = smac.validate(incumbent)\n    print(f\"Incumbent cost: {incumbent_cost}\")"
+        "from __future__ import annotations\n\nimport itertools\nimport warnings\n\nimport numpy as np\nfrom ConfigSpace import Categorical, Configuration, ConfigurationSpace, Float\nfrom sklearn import datasets\nfrom sklearn.linear_model import SGDClassifier\nfrom sklearn.model_selection import StratifiedKFold, cross_val_score\n\nfrom smac import MultiFidelityFacade as MFFacade\nfrom smac import Scenario\n\n__copyright__ = \"Copyright 2021, AutoML.org Freiburg-Hannover\"\n__license__ = \"3-clause BSD\"\n\n\nclass DigitsDataset:\n    def __init__(self) -> None:\n        self._data = datasets.load_digits()\n\n    def get_instances(self) -> list[str]:\n        \"\"\"Create instances from the dataset which include two classes only.\"\"\"\n        return [f\"{classA}-{classB}\" for classA, classB in itertools.combinations(self._data.target_names, 2)]\n\n    def get_instance_features(self) -> dict[str, list[int | float]]:\n        \"\"\"Returns the mean and variance of all instances as features.\"\"\"\n        features = {}\n        for instance in self.get_instances():\n            data, _ = self.get_instance_data(instance)\n            features[instance] = [np.mean(data), np.var(data)]\n\n        return features\n\n    def get_instance_data(self, instance: str) -> tuple[np.ndarray, np.ndarray]:\n        \"\"\"Retrieve data from the passed instance.\"\"\"\n        # We split the dataset into two classes\n        classA, classB = instance.split(\"-\")\n        indices = np.where(np.logical_or(int(classA) == self._data.target, int(classB) == self._data.target))\n\n        data = self._data.data[indices]\n        target = self._data.target[indices]\n\n        return data, target\n\n\nclass SGD:\n    def __init__(self, dataset: DigitsDataset) -> None:\n        self.dataset = dataset\n\n    @property\n    def configspace(self) -> ConfigurationSpace:\n        \"\"\"Build the configuration space which defines all parameters and their ranges for the SGD classifier.\"\"\"\n        cs = ConfigurationSpace()\n\n        # We define a few possible parameters for the SGD classifier\n        alpha = Float(\"alpha\", (0, 1), default=1.0)\n        l1_ratio = Float(\"l1_ratio\", (0, 1), default=0.5)\n        learning_rate = Categorical(\"learning_rate\", [\"constant\", \"invscaling\", \"adaptive\"], default=\"constant\")\n        eta0 = Float(\"eta0\", (0.00001, 1), default=0.1, log=True)\n        # Add the parameters to configuration space\n        cs.add_hyperparameters([alpha, l1_ratio, learning_rate, eta0])\n\n        return cs\n\n    def train(self, config: Configuration, instance: str, seed: int = 0) -> float:\n        \"\"\"Creates a SGD classifier based on a configuration and evaluates it on the\n        digits dataset using cross-validation.\"\"\"\n\n        with warnings.catch_warnings():\n            warnings.filterwarnings(\"ignore\")\n\n            # SGD classifier using given configuration\n            clf = SGDClassifier(\n                loss=\"log_loss\",\n                penalty=\"elasticnet\",\n                alpha=config[\"alpha\"],\n                l1_ratio=config[\"l1_ratio\"],\n                learning_rate=config[\"learning_rate\"],\n                eta0=config[\"eta0\"],\n                max_iter=30,\n                early_stopping=True,\n                random_state=seed,\n            )\n\n            # get instance\n            data, target = self.dataset.get_instance_data(instance)\n\n            cv = StratifiedKFold(n_splits=4, random_state=seed, shuffle=True)  # to make CV splits consistent\n            scores = cross_val_score(clf, data, target, cv=cv)\n\n        return 1 - np.mean(scores)\n\n\nif __name__ == \"__main__\":\n    dataset = DigitsDataset()\n    model = SGD(dataset)\n\n    scenario = Scenario(\n        model.configspace,\n        walltime_limit=30,  # We want to optimize for 30 seconds\n        n_trials=5000,  # We want to try max 5000 different trials\n        min_budget=1,  # Use min one instance\n        max_budget=45,  # Use max 45 instances (if we have a lot of instances we could constraint it here)\n        instances=dataset.get_instances(),\n        instance_features=dataset.get_instance_features(),\n    )\n\n    # Create our SMAC object and pass the scenario and the train method\n    smac = MFFacade(\n        scenario,\n        model.train,\n        overwrite=True,\n    )\n\n    # Now we start the optimization process\n    incumbent = smac.optimize()\n\n    default_cost = smac.validate(model.configspace.get_default_configuration())\n    print(f\"Default cost: {default_cost}\")\n\n    incumbent_cost = smac.validate(incumbent)\n    print(f\"Incumbent cost: {incumbent_cost}\")"
       ]
     }
   ],