CNNRegressor

Time Series Convolutional Neural Network (CNN), as described in [1].

Quickstart

from sktime.regression.deep_learning.cnn import CNNRegressor

estimator = CNNRegressor(n_epochs=2000, batch_size=16, kernel_size=7, avg_pool_size=3, n_conv_layers=2, callbacks=None, verbose=False, loss='mean_squared_error', metrics=None, random_state=0, activation='linear', activation_hidden='relu', use_bias=True, optimizer=None, filter_sizes=None, padding='auto')

Parameters(16)

n_epochsint, default = 2000

the number of epochs to train the model

batch_sizeint, default = 16

the number of samples per gradient update.

kernel_sizeint, default = 7

the length of the 1D convolution window

avg_pool_sizeint, default = 3

size of the average pooling windows

n_conv_layersint, default = 2

the number of convolutional plus average pooling layers

callbackslist of keras.callbacks, default = None

verboseboolean, default = False

whether to output extra information

lossstring, default=”mean_squared_error”

fit parameter for the keras model

metricslist of strings, default=[“accuracy”],

random_stateint or None, default=None

Seed for random number generation.

activationstring or a tf callable, default=”linear”

Activation function used in the output layer. List of available activation functions: https://keras.io/api/layers/activations/

activation_hiddenstring or a tf callable, default=”relu”

Activation function used in the hidden layers. List of available activation functions: https://keras.io/api/layers/activations/

use_biasboolean, default = True

whether the layer uses a bias vector.

optimizerkeras.optimizers object, default = Adam(lr=0.01)

specify the optimizer and the learning rate to be used.

filter_sizesarray of shape (n_conv_layers) default = [6, 12]

paddingstring, default = “auto”

Controls padding logic for the convolutional layers, i.e. whether 'valid' and 'same' are passed to the Conv1D layer. - “auto”: as per original implementation, "same" is passed if

input_shape[0] < 60 in the input layer, and "valid" otherwise.

• “valid”, “same”, and other values are passed directly to Conv1D

Examples

>>> from sktime.datasets import load_unit_test
>>> from sktime.regression.deep_learning.cnn import CNNRegressor
>>> X_train, y_train = load_unit_test (return_X_y = True, split = "train")
>>> X_test, y_test = load_unit_test (return_X_y = True, split = "test")
>>> regressor = CNNRegressor ()
>>> regressor. fit (X_train, y_train) CNNRegressor(
... )
>>> y_pred = regressor. predict (X_test)

References

1. [1 ] Zhao et. al, Convolutional neural networks for time series classification, Journal of Systems Engineering and Electronics, 28(1):2017.