Extending Auto-Sklearn with Classification Component

The following example demonstrates how to create a new classification component for using in auto-sklearn.

from typing import Optional
from pprint import pprint

from ConfigSpace.configuration_space import ConfigurationSpace
from ConfigSpace.hyperparameters import (
    CategoricalHyperparameter,
    UniformIntegerHyperparameter,
    UniformFloatHyperparameter,
)

import sklearn.metrics

from autosklearn.askl_typing import FEAT_TYPE_TYPE
import autosklearn.classification
import autosklearn.pipeline.components.classification
from autosklearn.pipeline.components.base import AutoSklearnClassificationAlgorithm
from autosklearn.pipeline.constants import (
    DENSE,
    SIGNED_DATA,
    UNSIGNED_DATA,
    PREDICTIONS,
)

from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split

Create MLP classifier component for auto-sklearn

class MLPClassifier(AutoSklearnClassificationAlgorithm):
    def __init__(
        self,
        hidden_layer_depth,
        num_nodes_per_layer,
        activation,
        alpha,
        solver,
        random_state=None,
    ):
        self.hidden_layer_depth = hidden_layer_depth
        self.num_nodes_per_layer = num_nodes_per_layer
        self.activation = activation
        self.alpha = alpha
        self.solver = solver
        self.random_state = random_state

    def fit(self, X, y):
        self.num_nodes_per_layer = int(self.num_nodes_per_layer)
        self.hidden_layer_depth = int(self.hidden_layer_depth)
        self.alpha = float(self.alpha)

        from sklearn.neural_network import MLPClassifier

        hidden_layer_sizes = tuple(
            self.num_nodes_per_layer for i in range(self.hidden_layer_depth)
        )

        self.estimator = MLPClassifier(
            hidden_layer_sizes=hidden_layer_sizes,
            activation=self.activation,
            alpha=self.alpha,
            solver=self.solver,
            random_state=self.random_state,
        )
        self.estimator.fit(X, y)
        return self

    def predict(self, X):
        if self.estimator is None:
            raise NotImplementedError()
        return self.estimator.predict(X)

    def predict_proba(self, X):
        if self.estimator is None:
            raise NotImplementedError()
        return self.estimator.predict_proba(X)

    @staticmethod
    def get_properties(dataset_properties=None):
        return {
            "shortname": "MLP Classifier",
            "name": "MLP CLassifier",
            "handles_regression": False,
            "handles_classification": True,
            "handles_multiclass": True,
            "handles_multilabel": False,
            "handles_multioutput": False,
            "is_deterministic": False,
            # Both input and output must be tuple(iterable)
            "input": [DENSE, SIGNED_DATA, UNSIGNED_DATA],
            "output": [PREDICTIONS],
        }

    @staticmethod
    def get_hyperparameter_search_space(
        feat_type: Optional[FEAT_TYPE_TYPE] = None, dataset_properties=None
    ):
        cs = ConfigurationSpace()
        hidden_layer_depth = UniformIntegerHyperparameter(
            name="hidden_layer_depth", lower=1, upper=3, default_value=1
        )
        num_nodes_per_layer = UniformIntegerHyperparameter(
            name="num_nodes_per_layer", lower=16, upper=216, default_value=32
        )
        activation = CategoricalHyperparameter(
            name="activation",
            choices=["identity", "logistic", "tanh", "relu"],
            default_value="relu",
        )
        alpha = UniformFloatHyperparameter(
            name="alpha", lower=0.0001, upper=1.0, default_value=0.0001
        )
        solver = CategoricalHyperparameter(
            name="solver", choices=["lbfgs", "sgd", "adam"], default_value="adam"
        )
        cs.add_hyperparameters(
            [
                hidden_layer_depth,
                num_nodes_per_layer,
                activation,
                alpha,
                solver,
            ]
        )
        return cs


# Add MLP classifier component to auto-sklearn.
autosklearn.pipeline.components.classification.add_classifier(MLPClassifier)
cs = MLPClassifier.get_hyperparameter_search_space()
print(cs)
Configuration space object:
  Hyperparameters:
    activation, Type: Categorical, Choices: {identity, logistic, tanh, relu}, Default: relu
    alpha, Type: UniformFloat, Range: [0.0001, 1.0], Default: 0.0001
    hidden_layer_depth, Type: UniformInteger, Range: [1, 3], Default: 1
    num_nodes_per_layer, Type: UniformInteger, Range: [16, 216], Default: 32
    solver, Type: Categorical, Choices: {lbfgs, sgd, adam}, Default: adam

Data Loading

X, y = load_breast_cancer(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(X, y)

Fit MLP classifier to the data

clf = autosklearn.classification.AutoSklearnClassifier(
    time_left_for_this_task=30,
    per_run_time_limit=10,
    include={"classifier": ["MLPClassifier"]},
    # Bellow two flags are provided to speed up calculations
    # Not recommended for a real implementation
    initial_configurations_via_metalearning=0,
    smac_scenario_args={"runcount_limit": 5},
)
clf.fit(X_train, y_train)
Fitting to the training data:   0%|          | 0/30 [00:00<?, ?it/s, The total time budget for this task is 0:00:30]
Fitting to the training data:   3%|3         | 1/30 [00:01<00:29,  1.00s/it, The total time budget for this task is 0:00:30]
Fitting to the training data:   7%|6         | 2/30 [00:02<00:28,  1.00s/it, The total time budget for this task is 0:00:30]
Fitting to the training data:  10%|#         | 3/30 [00:03<00:27,  1.00s/it, The total time budget for this task is 0:00:30]
Fitting to the training data:  13%|#3        | 4/30 [00:04<00:26,  1.00s/it, The total time budget for this task is 0:00:30]
Fitting to the training data:  17%|#6        | 5/30 [00:05<00:25,  1.00s/it, The total time budget for this task is 0:00:30]
Fitting to the training data:  20%|##        | 6/30 [00:06<00:24,  1.01s/it, The total time budget for this task is 0:00:30]
Fitting to the training data:  23%|##3       | 7/30 [00:07<00:23,  1.01s/it, The total time budget for this task is 0:00:30]
Fitting to the training data:  27%|##6       | 8/30 [00:08<00:22,  1.00s/it, The total time budget for this task is 0:00:30]
Fitting to the training data:  30%|###       | 9/30 [00:09<00:21,  1.00s/it, The total time budget for this task is 0:00:30]
Fitting to the training data:  33%|###3      | 10/30 [00:10<00:20,  1.00s/it, The total time budget for this task is 0:00:30]
Fitting to the training data:  37%|###6      | 11/30 [00:11<00:19,  1.00s/it, The total time budget for this task is 0:00:30]
Fitting to the training data:  40%|####      | 12/30 [00:12<00:18,  1.00s/it, The total time budget for this task is 0:00:30]
Fitting to the training data: 100%|##########| 30/30 [00:12<00:00,  2.49it/s, The total time budget for this task is 0:00:30]

AutoSklearnClassifier(ensemble_class=<class 'autosklearn.ensembles.ensemble_selection.EnsembleSelection'>,
                      include={'classifier': ['MLPClassifier']},
                      initial_configurations_via_metalearning=0,
                      per_run_time_limit=10,
                      smac_scenario_args={'runcount_limit': 5},
                      time_left_for_this_task=30)