[ENH] Migrate ResNetClassifier from sktime-dl to sktime (#3881)

Part of #3351. See also #3365 

Migrates `ResNetClassifier` from `sktime-dl` to `sktime`

.all-contributorsrc

@@ -156,6 +156,15 @@
         "design"
       ]
     },
+    {
+      "login": "nilesh05apr",
+      "name": "Nilesh Kumar",
+      "avatar_url": "https://avatars.githubusercontent.com/u/65773314?v=4",
+      "profile": "https://github.com/nilesh05apr"
+      "contributions": [
+      "code",
+      ]
+    },
     {
       "login": "MatthewMiddlehurst",
       "name": "Matthew Middlehurst",

sktime/classification/deep_learning/resnet.py

@@ -0,0 +1,215 @@
+# -*- coding: utf-8 -*-
+"""Residual Network (ResNet) for classification."""
+
+__author__ = ["James-Large", "AurumnPegasus", "nilesh05apr"]
+__all__ = ["ResNetClassifier"]
+
+from sklearn.utils import check_random_state
+
+from sktime.classification.deep_learning.base import BaseDeepClassifier
+from sktime.networks.resnet import ResNetNetwork
+from sktime.utils.validation._dependencies import _check_dl_dependencies
+
+_check_dl_dependencies(severity="warning")
+
+
+class ResNetClassifier(BaseDeepClassifier):
+    """
+    Residual Neural Network as described in [1].
+
+    Parameters
+    ----------
+    n_epochs       : int, default = 1500
+        the number of epochs to train the model
+    batch_size      : int, default = 16
+        the number of samples per gradient update.
+    random_state    : int or None, default=None
+        Seed for random number generation.
+    verbose         : boolean, default = False
+        whether to output extra information
+    loss            : string, default="mean_squared_error"
+        fit parameter for the keras model
+    optimizer       : keras.optimizer, default=keras.optimizers.Adam(),
+    metrics         : list of strings, default=["accuracy"],
+    activation      : string or a tf callable, default="sigmoid"
+        Activation function used in the output linear layer.
+        List of available activation functions:
+        https://keras.io/api/layers/activations/
+    use_bias        : boolean, default = True
+        whether the layer uses a bias vector.
+    optimizer       : keras.optimizers object, default = Adam(lr=0.01)
+        specify the optimizer and the learning rate to be used.
+
+
+    Notes
+    -----
+    Adapted from the implementation from source code
+    https://github.com/hfawaz/dl-4-tsc/blob/master/classifiers/resnet.py
+
+    References
+    ----------
+        .. [1] Wang et. al, Time series classification from
+    scratch with deep neural networks: A strong baseline,
+    International joint conference on neural networks (IJCNN), 2017.
+
+    Examples
+    --------
+    >>> from sktime.classification.deep_learning.resnet import ResNetClassifier
+    >>> from sktime.datasets import load_unit_test
+    >>> X_train, y_train = load_unit_test(split="train")
+    >>> clf = ResNetClassifier(n_epochs=20, bacth_size=4) # doctest: +SKIP
+    >>> clf.fit(X_train, Y_train) # doctest: +SKIP
+    ResNetClassifier(...)
+    """
+
+    _tags = {"python_dependencies": ["tensorflow"]}
+
+    def __init__(
+        self,
+        n_epochs=1500,
+        callbacks=None,
+        verbose=False,
+        loss="categorical_crossentropy",
+        metrics=None,
+        batch_size=16,
+        random_state=None,
+        activation="sigmoid",
+        use_bias=True,
+        optimizer=None,
+    ):
+        _check_dl_dependencies(severity="error")
+        super(ResNetClassifier, self).__init__()
+        self.n_epochs = n_epochs
+        self.callbacks = callbacks
+        self.verbose = verbose
+        self.loss = loss
+        self.metrics = metrics
+        self.batch_size = batch_size
+        self.random_state = random_state
+        self.activation = activation
+        self.use_bias = use_bias
+        self.optimizer = optimizer
+        self.history = None
+        self._network = ResNetNetwork(random_state=random_state)
+
+    def build_model(self, input_shape, n_classes, **kwargs):
+        """Construct a compiled, un-trained, keras model that is ready for training.
+
+        In sktime, time series are stored in numpy arrays of shape (d,m), where d
+        is the number of dimensions, m is the series length. Keras/tensorflow assume
+        data is in shape (m,d). This method also assumes (m,d). Transpose should
+        happen in fit.
+
+        Parameters
+        ----------
+        input_shape : tuple
+            The shape of the data fed into the input layer, should be (m,d)
+        n_classes: int
+            The number of classes, which becomes the size of the output layer
+
+        Returns
+        -------
+        output : a compiled Keras Model
+        """
+        import tensorflow as tf
+        from tensorflow import keras
+
+        tf.random.set_seed(self.random_state)
+
+        self.optimizer_ = (
+            keras.optimizers.Adam(learning_rate=0.01)
+            if self.optimizer is None
+            else self.optimizer
+        )
+
+        if self.metrics is None:
+            metrics = ["accuracy"]
+        else:
+            metrics = self.metrics
+
+        input_layer, output_layer = self._network.build_network(input_shape, **kwargs)
+
+        output_layer = keras.layers.Dense(
+            units=n_classes, activation=self.activation, use_bias=self.use_bias
+        )(output_layer)
+
+        model = keras.models.Model(inputs=input_layer, outputs=output_layer)
+        model.compile(
+            loss=self.loss,
+            optimizer=self.optimizer_,
+            metrics=metrics,
+        )
+
+        return model
+
+    def _fit(self, X, y):
+        """Fit the classifier on the training set (X, y).
+
+        Parameters
+        ----------
+        X : np.ndarray of shape = (n_instances (n), n_dimensions (d), series_length (m))
+            The training input samples.
+        y : np.ndarray of shape n
+            The training data class labels.
+
+        Returns
+        -------
+        self : object
+        """
+        if self.callbacks is None:
+            self._callbacks = []
+
+        y_onehot = self.convert_y_to_keras(y)
+        # Transpose to conform to Keras input style.
+        X = X.transpose(0, 2, 1)
+
+        check_random_state(self.random_state)
+        self.input_shape = X.shape[1:]
+        self.model_ = self.build_model(self.input_shape, self.n_classes_)
+        if self.verbose:
+            self.model_.summary()
+        self.history = self.model_.fit(
+            X,
+            y_onehot,
+            batch_size=self.batch_size,
+            epochs=self.n_epochs,
+            verbose=self.verbose,
+            callbacks=self._callbacks,
+        )
+        return self
+
+    @classmethod
+    def get_test_params(cls, parameter_set="default"):
+        """Return testing parameter settings for the estimator.
+
+        Parameters
+        ----------
+        parameter_set : str, default="default"
+            Name of the set of test parameters to return, for use in tests. If no
+            special parameters are defined for a value, will return `"default"` set.
+            For classifiers, a "default" set of parameters should be provided for
+            general testing, and a "results_comparison" set for comparing against
+            previously recorded results if the general set does not produce suitable
+            probabilities to compare against.
+
+        Returns
+        -------
+        params : dict or list of dict, default={}
+            Parameters to create testing instances of the class.
+            Each dict are parameters to construct an "interesting" test instance, i.e.,
+            `MyClass(**params)` or `MyClass(**params[i])` creates a valid test instance.
+            `create_test_instance` uses the first (or only) dictionary in `params`.
+        """
+        param1 = {
+            "n_epochs": 10,
+            "batch_size": 4,
+            "use_bias": False,
+        }
+
+        param2 = {
+            "n_epochs": 12,
+            "batch_size": 6,
+            "use_bias": True,
+        }
+
+        return [param1, param2]

sktime/networks/resnet.py

@@ -0,0 +1,154 @@
+# -*- coding: utf-8 -*-
+"""Residual Network (ResNet) (minus the final output layer)."""
+
+__author__ = ["James Large", "Withington", "nilesh05apr"]
+
+from sktime.networks.base import BaseDeepNetwork
+from sktime.utils.validation._dependencies import _check_dl_dependencies
+
+
+class ResNetNetwork(BaseDeepNetwork):
+    """
+    Establish the network structure for a ResNet.
+
+    Adapted from the implementations used in [1]
+
+    Parameters
+    ----------
+    random_state : int, optional (default = 0)
+        The random seed to use random activities.
+
+    Notes
+    -----
+    Adpated from the implementation source code
+    https://github.com/hfawaz/dl-4-tsc/blob/master/classifiers/resnet.py
+
+    References
+    ----------
+    .. [1] H. Fawaz, G. B. Lanckriet, F. Petitjean, and L. Idoumghar,
+
+    Network originally defined in:
+
+    @inproceedings{wang2017time, title={Time series classification from
+    scratch with deep neural networks: A strong baseline}, author={Wang,
+    Zhiguang and Yan, Weizhong and Oates, Tim}, booktitle={2017
+    International joint conference on neural networks (IJCNN)}, pages={
+    1578--1585}, year={2017}, organization={IEEE} }
+
+    """
+
+    _tags = {"python_dependencies": ["tensorflow", "keras-self-attention"]}
+
+    def __init__(self, random_state=0):
+        _check_dl_dependencies(severity="error")
+        super(ResNetNetwork, self).__init__()
+        self.random_state = random_state
+
+    def build_network(self, input_shape, **kwargs):
+        """
+        Construct a network and return its input and output layers.
+
+        Arguments
+        ---------
+        input_shape : tuple of shape = (series_length (m), n_dimensions (d))
+            The shape of the data fed into the input layer.
+
+        Returns
+        -------
+        input_layer : keras.layers.Input
+            The input layer of the network.
+        output_layer : keras.layers.Layer
+            The output layer of the network.
+        """
+        from tensorflow import keras
+
+        n_feature_maps = 64
+
+        input_layer = keras.layers.Input(input_shape)
+
+        # 1st residual block
+
+        conv_x = keras.layers.Conv1D(
+            filters=n_feature_maps, kernel_size=8, padding="same"
+        )(input_layer)
+        conv_x = keras.layers.BatchNormalization()(conv_x)
+        conv_x = keras.layers.Activation("relu")(conv_x)
+
+        conv_y = keras.layers.Conv1D(
+            filters=n_feature_maps, kernel_size=5, padding="same"
+        )(conv_x)
+        conv_y = keras.layers.BatchNormalization()(conv_y)
+        conv_y = keras.layers.Activation("relu")(conv_y)
+
+        conv_z = keras.layers.Conv1D(
+            filters=n_feature_maps, kernel_size=3, padding="same"
+        )(conv_y)
+        conv_z = keras.layers.BatchNormalization()(conv_z)
+
+        # expand channels for the sum
+        shortcut_y = keras.layers.Conv1D(
+            filters=n_feature_maps, kernel_size=1, padding="same"
+        )(input_layer)
+        shortcut_y = keras.layers.BatchNormalization()(shortcut_y)
+
+        output_block_1 = keras.layers.add([shortcut_y, conv_z])
+        output_block_1 = keras.layers.Activation("relu")(output_block_1)
+
+        # 2nd residual block
+
+        conv_x = keras.layers.Conv1D(
+            filters=n_feature_maps * 2, kernel_size=8, padding="same"
+        )(output_block_1)
+        conv_x = keras.layers.BatchNormalization()(conv_x)
+        conv_x = keras.layers.Activation("relu")(conv_x)
+
+        conv_y = keras.layers.Conv1D(
+            filters=n_feature_maps * 2, kernel_size=5, padding="same"
+        )(conv_x)
+        conv_y = keras.layers.BatchNormalization()(conv_y)
+        conv_y = keras.layers.Activation("relu")(conv_y)
+
+        conv_z = keras.layers.Conv1D(
+            filters=n_feature_maps * 2, kernel_size=3, padding="same"
+        )(conv_y)
+        conv_z = keras.layers.BatchNormalization()(conv_z)
+
+        # expand channels for the sum
+        shortcut_y = keras.layers.Conv1D(
+            filters=n_feature_maps * 2, kernel_size=1, padding="same"
+        )(output_block_1)
+        shortcut_y = keras.layers.BatchNormalization()(shortcut_y)
+
+        output_block_2 = keras.layers.add([shortcut_y, conv_z])
+        output_block_2 = keras.layers.Activation("relu")(output_block_2)
+
+        # 3rd residual block
+
+        conv_x = keras.layers.Conv1D(
+            filters=n_feature_maps * 2, kernel_size=8, padding="same"
+        )(output_block_2)
+        conv_x = keras.layers.BatchNormalization()(conv_x)
+        conv_x = keras.layers.Activation("relu")(conv_x)
+
+        conv_y = keras.layers.Conv1D(
+            filters=n_feature_maps * 2, kernel_size=5, padding="same"
+        )(conv_x)
+        conv_y = keras.layers.BatchNormalization()(conv_y)
+        conv_y = keras.layers.Activation("relu")(conv_y)
+
+        conv_z = keras.layers.Conv1D(
+            filters=n_feature_maps * 2, kernel_size=3, padding="same"
+        )(conv_y)
+        conv_z = keras.layers.BatchNormalization()(conv_z)
+
+        # no need to expand channels because they are equal
+        shortcut_y = keras.layers.BatchNormalization()(output_block_2)
+
+        output_block_3 = keras.layers.add([shortcut_y, conv_z])
+        output_block_3 = keras.layers.Activation("relu")(output_block_3)
+
+        # global average pooling
+
+        gap_layer = keras.layers.GlobalAveragePooling1D()(output_block_3)
+
+        return input_layer, gap_layer

sktime/tests/_config.py

@@ -77,6 +77,9 @@
         "test_save_estimators_to_file",
     ],
     # `test_fit_idempotent` fails with `AssertionError`, see #3616
+    "ResNetClassifier": [
+        "test_fit_idempotent",
+    ],
     "CNNClassifier": [
         "test_fit_idempotent",
     ],