TapNetRegressor

TapNetRegressor

class TapNetRegressor(n_epochs=2000, batch_size=16, dropout=0.5, filter_sizes=(256, 256, 128), kernel_size=(8, 5, 3), dilation=1, layers=(500, 300), use_rp=True, activation='linear', activation_hidden='leaky_relu', rp_group=3, rp_alpha=2.0, use_bias=True, use_att=True, use_lstm=True, use_cnn=True, random_state=None, padding='same', loss='mean_squared_error', optimizer=None, metrics=None, callbacks=None, verbose=False, lstm_dropout=0.8, fc_dropout=0.0)[source]

Time series attentional prototype network (TapNet), as described in [1].

TapNet was initially proposed for multivariate time series classification. This an adaptation for time series regression. TapNet comprises these components: random dimension permutation, multivariate time series encoding, and attentional prototype learning.

Parameters:

n_epochsint, default = 2000

number of epochs to train the model

batch_sizeint, default = 16

number of samples per update

dropoutfloat, default = 0.5

dropout rate for the convolutional layers, in the range [0, 1)

filter_sizestuple of int, default = (256, 256, 128)

number of convolutional filters in each conv block

kernel_sizetuple of int, default = (8, 5, 3)

specifying the length of the 1D convolution window

dilationint, default = 1

dilation value

layersarray of int, default = (500, 300)

sizes of dense layers

use_rpbool, default = True

whether to use random projections

activationstr or callable, default = “linear”

activation function for the last output layer List of available activation functions: https://keras.io/api/layers/activations/

activation_hiddenstr or callable, default = “leaky_relu”

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

rp_groupint, default = 3

number of random permutation groups g for random dimension permutation

rp_alphafloat, default = 2.0

scale factor alpha used to compute the RDP group size: ` rp_dim = floor(n_dims * rp_alpha / rp_group) `

use_biasbool, default = True

whether to use bias in the output dense layer

use_attbool, default = True

whether to use self attention

use_lstmbool, default = True

whether to use an LSTM layer

use_cnnbool, default = True

whether to use a CNN layer

random_stateint or None, default = None

seed to any needed random actions

paddingstr, default = ‘same’

type of padding for convolution layers

lossstr, default = “mean_squared_error”

loss function for the classifier

optimizerstr or None, default = “Adam(lr=0.01)”

gradient updating function for the classifier

metricslist of str or None, default = None

list of metrics to be evaluated by the model during training and testing

callbackslist of keras.callbacks.Callback, optional (default=None)

List of Keras callbacks to apply during model training.

verbosebool, default = False

whether to output extra information

lstm_dropoutfloat, default = 0.8

dropout rate for the LSTM layer, in the range [0, 1)

fc_dropoutfloat, default = 0.0

dropout rate before the output layer, in the range (0, 1]

Attributes:

is_fitted

Whether fit has been called.

Notes

The Implementation of TapNet found at https://github.com/kdd2019-tapnet/tapnet Currently does not implement custom distance matrix loss function or class based self attention.

References

[1]

Zhang et al. Tapnet: Multivariate time series classification with

attentional prototypical network, Proceedings of the AAAI Conference on Artificial Intelligence 34(4), 6845-6852, 2020

Examples

>>> from sktime.regression.deep_learning.tapnet import TapNetRegressor
>>> from sktime.datasets import load_unit_test
>>> X_train, y_train = load_unit_test(split="train")
>>> X_test, y_test = load_unit_test(split="test")
>>> reg = TapNetRegressor(n_epochs=20, batch_size=4)
>>> reg.fit(X_train, y_train)
TapNetRegressor(...)

Methods

build_model(input_shape, **kwargs)	Construct a complied, un-trained, keras model that is ready for training.
check_is_fitted([method_name])	Check if the estimator has been fitted.
clone()	Obtain a clone of the object with same hyper-parameters and config.
clone_tags(estimator[, tag_names])	Clone tags from another object as dynamic override.
create_test_instance([parameter_set])	Construct an instance of the class, using first test parameter set.
create_test_instances_and_names([parameter_set])	Create list of all test instances and a list of names for them.
fit(X, y)	Fit time series regressor to training data.
get_class_tag(tag_name[, tag_value_default])	Get class tag value from class, with tag level inheritance from parents.
get_class_tags()	Get class tags from class, with tag level inheritance from parent classes.
get_config()	Get config flags for self.
get_fitted_params([deep])	Get fitted parameters.
get_param_defaults()	Get object's parameter defaults.
get_param_names([sort])	Get object's parameter names.
get_params([deep])	Get a dict of parameters values for this object.
get_tag(tag_name[, tag_value_default, ...])	Get tag value from instance, with tag level inheritance and overrides.
get_tags()	Get tags from instance, with tag level inheritance and overrides.
get_test_params([parameter_set])	Return testing parameter settings for the estimator.
is_composite()	Check if the object is composed of other BaseObjects.
load_from_path(serial)	Load object from file location.
load_from_serial(serial)	Load object from serialized memory container.
predict(X)	Predicts labels for sequences in X.
reset()	Reset the object to a clean post-init state.
save([path])	Save serialized self to bytes-like object or to (.zip) file.
score(X, y[, multioutput])	Scores predicted labels against ground truth labels on X.
set_config(**config_dict)	Set config flags to given values.
set_params(**params)	Set the parameters of this object.
set_random_state([random_state, deep, ...])	Set random_state pseudo-random seed parameters for self.
set_tags(**tag_dict)	Set instance level tag overrides to given values.