ResNetNetwork

class ResNetNetwork(n_residual_blocks=3, n_conv_per_residual_block=3, n_filters=None, kernel_size=None, strides=1, dilation_rate=1, padding='same', activation='relu', use_bias=True)

Establish the network structure for a ResNet.

Adapted from the implementations used in [1].

Parameters:

n_residual_blocksint, default = 3

The number of residual blocks of ResNet’s model.

n_conv_per_residual_blockint, default = 3

The number of convolution blocks in each residual block.

n_filtersint or list of int, default = [128, 64, 64]

The number of convolution filters for all the convolution layers in the same residual block, if not a list, the same number of filters is used in all convolutions of all residual blocks.

kernel_sizeint or list of int, default = [8, 5, 3]

The kernel size of all the convolution layers in one residual block, if not a list, the same kernel size is used in all convolution layers.

stridesint or list of int, default = 1

The strides of convolution kernels in each of the convolution layers in one residual block, if not a list, the same kernel size is used in all convolution layers.

dilation_rateint or list of int, default = 1

The dilation rate of the convolution layers in one residual block, if not a list, the same kernel size is used in all convolution layers.

paddingstr or list of str, default = ‘padding’

The type of padding used in the convolution layers in one residual block, if not a list, the same kernel size is used in all convolution layers.

activationstr or list of str, default = ‘relu’

Keras activation used in the convolution layers in one residual block, if not a list, the same kernel size is used in all convolution layers.

use_biasbool or list of bool, default = True

Condition on whether or not to use bias values in the convolution layers in one residual block, if not a list, the same kernel size is used in all convolution layers.

Notes

Adpated from the implementation source code https://github.com/hfawaz/dl-4-tsc/blob/master/classifiers/resnet.py

References

[1]

8. 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} }

Methods

build_network(input_shape, **kwargs)

Construct a network and return its input and output layers.

build_network(input_shape, **kwargs)

Construct a network and return its input and output layers.

Parameters:

input_shapetuple of shape = (n_timepoints (m), n_channels (d))

The shape of the data fed into the input layer.

Returns:

input_layerkeras.layers.Input

The input layer of the network.

output_layerkeras.layers.Layer

The output layer of the network.