[ENH] Added rocket on gpu

aeon/transformations/collection/convolution_based/rocketGPU/__init__.py

@@ -0,0 +1,7 @@
+"""Rocket transformers for GPU."""
+
+__all__ = ["ROCKETGPU"]
+
+from aeon.transformations.collection.convolution_based.rocketGPU._rocket_gpu import (
+    ROCKETGPU,
+)

aeon/transformations/collection/convolution_based/rocketGPU/_rocket_gpu.py

@@ -0,0 +1,230 @@
+"""Rocket transformer for GPU."""
+
+__author__ = ["hadifawaz1999"]
+__all__ = ["ROCKETGPU"]
+
+from aeon.transformations.collection.convolution_based.rocketGPU.base import (
+    BaseROCKETGPU,
+)
+
+
+class ROCKETGPU(BaseROCKETGPU):
+    """RandOm Convolutional KErnel Transform (ROCKET) for GPU.
+
+    A kernel (or convolution) is a subseries used to create features that can be used
+    in machine learning tasks. ROCKET [1]_ generates a large number of random
+    convolutional kernels in the fit method. The length and dilation of each kernel
+    are also randomly generated. The kernels are used in the transform stage to
+    generate a new set of features. A kernel is used to create an activation map for
+    each series by running it across a time series, including random length and
+    dilation. It transforms the time series with two features per kernel. The first
+    feature is global max pooling and the second is proportion of positive values
+    (or PPV).
+
+
+    Parameters
+    ----------
+    n_filters : int, default=10,000
+       Number of random convolutional filters.
+    kernel_size : list, default = None
+        The list of possible kernel sizes, default is [7, 9, 11].
+    padding : list, default = None
+        The list of possible tensorflow padding, default is ["SAME","VALID"].
+    use_dilation : bool, default = True
+        Whether or not to use dilation in convolution operations.
+    bias_range : Tuple, default = None
+        The min and max value of bias values, default is (-1.0, 1.0).
+    batch_size : int, default = 64
+        The batch to parallelize over GPU.
+    random_state : None or int, optional, default = None
+        Seed for random number generation.
+
+    References
+    ----------
+    .. [1] Tan, Chang Wei and Dempster, Angus and Bergmeir, Christoph
+        and Webb, Geoffrey I,
+        "ROCKET: Exceptionally fast and accurate time series
+      classification using random convolutional kernels",2020,
+      https://link.springer.com/article/10.1007/s10618-020-00701-z,
+      https://arxiv.org/abs/1910.13051
+    """
+
+    def __init__(
+        self,
+        n_filters=1000,
+        kernel_size=None,
+        padding=None,
+        use_dilation=True,
+        bias_range=None,
+        batch_size=64,
+        random_state=None,
+    ):
+        super().__init__(n_filters)
+
+        self.n_filters = n_filters
+        self.kernel_size = kernel_size
+        self.padding = padding
+        self.use_dilation = use_dilation
+        self.bias_range = bias_range
+        self.batch_size = batch_size
+        self.random_state = random_state
+
+    def _define_parameters(self):
+        """Define the parameters of ROCKET."""
+        import numpy as np
+
+        np.random.seed(self.random_state)
+
+        self._list_of_kernels = []
+        self._list_of_dilations = []
+        self._list_of_paddings = []
+        self._list_of_biases = []
+
+        for _ in range(self.n_filters):
+            _kernel_size = np.random.choice(self._kernel_size, size=1)[0]
+            _convolution_kernel = np.random.normal(
+                size=(_kernel_size, self.n_channels, 1)
+            )
+            _convolution_kernel = _convolution_kernel - _convolution_kernel.mean(
+                axis=0, keepdims=True
+            )
+
+            if self.use_dilation:
+                _dilation_rate = 2 ** np.random.uniform(
+                    0, np.log2((self.input_length - 1) / (_kernel_size - 1))
+                )
+            else:
+                _dilation_rate = 1
+
+            _padding = np.random.choice(self._padding, size=1)[0]
+            assert _padding in ["SAME", "VALID"]
+
+            _bias = np.random.uniform(self._bias_range[0], self._bias_range[1])
+
+            self._list_of_kernels.append(_convolution_kernel)
+            self._list_of_dilations.append(_dilation_rate)
+            self._list_of_paddings.append(_padding)
+            self._list_of_biases.append(_bias)
+
+    def _fit(self, X, y=None):
+        """Generate random kernels adjusted to time series shape.
+
+        Infers time series length and number of channels from input numpy array,
+        and generates random kernels.
+
+        Parameters
+        ----------
+        X : 3D np.ndarray of shape = (n_instances, n_channels, n_timepoints)
+            collection of time series to transform.
+        y : ignored argument for interface compatibility.
+
+        Returns
+        -------
+        self
+        """
+        self.input_length = X.shape[2]
+        self.n_channels = X.shape[1]
+
+        self._kernel_size = [7, 9, 11] if self.kernel_size is None else self.kernel_size
+        self._padding = ["VALID", "SAME"] if self.padding is None else self.padding
+        self._bias_range = (-1.0, 1.0) if self.bias_range is None else self.bias_range
+
+        assert self._bias_range[0] <= self._bias_range[1]
+
+        self._define_parameters()
+
+    def _generate_batch_indices(self, n):
+        """Generate the list of batches.
+
+        Parameters
+        ----------
+        n : int
+            The number of samples in the dataset.
+
+        Returns
+        -------
+        batch_indices_list : list
+            A list of multiple np.ndarray containing indices of batches.
+        """
+        import numpy as np
+
+        all_indices = np.arange(n)
+
+        if self.batch_size >= n:
+            return [all_indices]
+
+        remainder_batch_size = n % self.batch_size
+        number_batches = n // self.batch_size
+
+        batch_indices_list = np.array_split(
+            ary=all_indices[: n - remainder_batch_size],
+            indices_or_sections=number_batches,
+        )
+
+        if remainder_batch_size > 0:
+            batch_indices_list.append(all_indices[n - remainder_batch_size :])
+
+        return batch_indices_list
+
+    def _transform(self, X, y=None):
+        """Transform input time series using random convolutional kernels.
+
+        Parameters
+        ----------
+        X : 3D np.ndarray of shape = [n_instances, n_channels, n_timepoints]
+            collection of time series to transform.
+        y : ignored argument for interface compatibility.
+
+        Returns
+        -------
+        output_rocket : np.ndarray [n_instances, n_filters * 2]
+            transformed features.
+        """
+        import numpy as np
+        import tensorflow as tf
+
+        tf.random.set_seed(self.random_state)
+
+        X = X.transpose(0, 2, 1)
+
+        batch_indices_list = self._generate_batch_indices(n=len(X))
+
+        output_features = []
+
+        for f in range(self.n_filters):
+            output_features_filter = []
+
+            for batch_indices in batch_indices_list:
+                _output_convolution = tf.nn.conv1d(
+                    input=X[batch_indices],
+                    stride=1,
+                    filters=self._list_of_kernels[f],
+                    dilations=self._list_of_dilations[f],
+                    padding=self._list_of_paddings[f],
+                )
+
+                _output_convolution = _output_convolution.numpy().reshape(
+                    (len(batch_indices), len(_output_convolution[0]))
+                )
+                _output_convolution += self._list_of_biases[f]
+
+                _ppv = self._get_ppv(x=_output_convolution)
+                _max = self._get_max(x=_output_convolution)
+
+                output_features_filter.append(
+                    np.concatenate(
+                        (np.expand_dims(_ppv, axis=-1), np.expand_dims(_max, axis=-1)),
+                        axis=1,
+                    )
+                )
+
+            output_features.append(np.concatenate(output_features_filter, axis=0))
+
+        output_rocket = (
+            np.concatenate(output_features, axis=0)
+            .reshape(self.n_filters, len(X), 2)
+            .transpose([1, 0, 2])
+        )
+        output_rocket = output_rocket.reshape((len(X), self.n_filters * 2))
+
+        return output_rocket

aeon/transformations/collection/convolution_based/rocketGPU/base.py

@@ -0,0 +1,44 @@
+"""Base of Rocket based transformer for GPU."""
+
+__author__ = ["hadifawaz1999"]
+__all__ = ["BaseROCKETGPU"]
+
+from aeon.transformations.collection import BaseCollectionTransformer
+
+
+class BaseROCKETGPU(BaseCollectionTransformer):
+    """Base class for ROCKET GPU based transformers.
+
+    Parameters
+    ----------
+    nb_filters : int, default = 1000
+        Number of random convolutional kernels.
+    """
+
+    _tags = {
+        "X_inner_type": "numpy3D",
+        "output_data_type": "Tabular",
+        "capability:multivariate": True,
+        "algorithm_type": "convolution",
+        "capability:unequal_length": False,
+        "cant-pickle": True,
+        "python_dependencies": "tensorflow",
+    }
+
+    def __init__(
+        self,
+        n_filters=10000,
+    ):
+        super().__init__()
+        self.n_filters = n_filters
+
+    def _get_ppv(self, x):
+        import tensorflow as tf
+
+        x_pos = tf.math.count_nonzero(tf.nn.relu(x), axis=1)
+        return tf.math.divide(x_pos, x.shape[1])
+
+    def _get_max(self, x):
+        import tensorflow as tf
+
+        return tf.math.reduce_max(x, axis=1)