create_architecture {cito}

CNN architecture

Description

creates a 'citoarchitecture' object that is used by cnn.

Usage

create_architecture(
  ...,
  default_n_neurons = 10,
  default_n_kernels = 10,
  default_kernel_size = list(conv = 3, maxPool = 2, avgPool = 2),
  default_stride = list(conv = 1, maxPool = NULL, avgPool = NULL),
  default_padding = list(conv = 0, maxPool = 0, avgPool = 0),
  default_dilation = list(conv = 1, maxPool = 1),
  default_bias = list(conv = TRUE, linear = TRUE),
  default_activation = list(conv = "relu", linear = "relu"),
  default_normalization = list(conv = FALSE, linear = FALSE),
  default_dropout = list(conv = 0, linear = 0)
)

Arguments

...

objects of class 'citolayer' created by linear, conv, maxPool, avgPool or transfer

default_n_neurons

(int) default value: amount of neurons in a linear layer

default_n_kernels

(int) default value: amount of kernels in a convolutional layer

default_kernel_size

(int or tuple) default value: size of the kernels in convolutional and pooling layers. Use a tuple if the kernel size isn't equal in all dimensions

default_stride

(int or tuple) default value: stride of the kernels in convolutional and pooling layers. NULL sets the stride equal to the kernel size. Use a tuple if the stride isn't equal in all dimensions

default_padding

(int or tuple) default value: zero-padding added to both sides of the input. Use a tuple if the padding isn't equal in all dimensions

default_dilation

(int or tuple) default value: dilation of the kernels in convolutional and maxPooling layers. Use a tuple if the dilation isn't equal in all dimensions

default_bias

(boolean) default value: if TRUE, adds a learnable bias to neurons of linear and kernels of convolutional layers

default_activation

(string) default value: activation function that is used after linear and convolutional layers. The following activation functions are supported: "relu", "leaky_relu", "tanh", "elu", "rrelu", "prelu", "softplus", "celu", "selu", "gelu", "relu6", "sigmoid", "softsign", "hardtanh", "tanhshrink", "softshrink", "hardshrink", "log_sigmoid"

default_normalization

(boolean) default value: if TRUE, batch normalization is used after linear and convolutional layers

default_dropout

(float) default value: dropout rate of linear and convolutional layers. Set to 0 for no dropout

Details

This function creates a 'citoarchitecture' object that provides the cnn function with all information about the architecture of the CNN that will be created and trained. The final architecture consists of the layers in the sequence they were passed to this function. All parameters of the 'citolayer' objects, that are still NULL because they haven't been specified at the creation of the layer, are filled with the given default parameters for their specific layer type (linear, conv, maxPool, avgPool). The default values can be changed by either passing a list with the values for specific layer types (in which case the defaults of layer types which aren't in the list remain the same) or by passing a single value (in which case the defaults for all layer types is set to that value).

Value

S3 object of class "citoarchitecture"

Author(s)

Armin Schenk

See Also

cnn, linear, conv, maxPool, avgPool, transfer, print.citoarchitecture, plot.citoarchitecture