% Generated by roxygen2: do not edit by hand
% Please edit documentation in R/create_model_genomenet.R
\name{create_model_genomenet}
\alias{create_model_genomenet}
\title{Create GenomeNet Model with Given Architecture Parameters}
\usage{
create_model_genomenet(
  maxlen = 300,
  learning_rate = 0.001,
  number_of_cnn_layers = 1,
  conv_block_count = 1,
  kernel_size_0 = 16,
  kernel_size_end = 16,
  filters_0 = 256,
  filters_end = 512,
  dilation_end = 1,
  max_pool_end = 1,
  dense_layer_num = 1,
  dense_layer_units = 100,
  dropout_lstm = 0,
  dropout = 0,
  batch_norm_momentum = 0.8,
  leaky_relu_alpha = 0,
  dense_activation = "relu",
  skip_block_fraction = 0,
  residual_block = FALSE,
  reverse_encoding = FALSE,
  optimizer = "adam",
  model_type = "gap",
  recurrent_type = "lstm",
  recurrent_layers = 1,
  recurrent_bidirectional = FALSE,
  recurrent_units = 100,
  vocabulary_size = 4,
  last_layer_activation = "softmax",
  loss_fn = "categorical_crossentropy",
  auc_metric = FALSE,
  num_targets = 2,
  model_seed = NULL,
  bal_acc = FALSE,
  f1_metric = FALSE,
  mixed_precision = FALSE,
  mirrored_strategy = NULL
)
}
\arguments{
\item{maxlen}{(integer \code{numeric(1)})\cr
Input sequence length.}

\item{learning_rate}{(\code{numeric(1)})\cr
Used by the \code{keras} optimizer that is specified by \code{optimizer}.}

\item{number_of_cnn_layers}{(integer \code{numeric(1)})\cr
Target number of CNN-layers to use in total. If \code{number_of_cnn_layers} is
greater than \code{conv_block_count}, then the effective number of CNN layers
is set to the closest integer that is divisible by \code{conv_block_count}.}

\item{conv_block_count}{(integer \code{numeric(1)})\cr
Number of convolutional blocks, into which the CNN layers are divided.
If this is greater than \code{number_of_cnn_layers}, then it is set to
\code{number_of_cnn_layers} (the convolutional block size will then be 1).\cr
Convolutional blocks are used when \code{model_type} is \code{"gap"} (the output of
the last \code{conv_block_count * (1 - skip_block_fraction)} blocks is
fed to global average pooling and then concatenated), and also when
\code{residual_block} is \code{TRUE} (the number of filters is held constant within
blocks). If neither of these is the case, \code{conv_block_count} has little
effect besides the fact that \code{number_of_cnn_layers} is set to the closest
integer divisible by \code{conv_block_count}.}

\item{kernel_size_0}{(\code{numeric(1)})\cr
Target CNN kernel size of the first CNN-layer. Although CNN kernel size is
always an integer, this value can be non-integer, potentially affecting
the kernel-sizes of intermediate layers (which are geometrically
interpolated between \code{kernel_size_0} and \code{kernel_size_end}).}

\item{kernel_size_end}{(\code{numeric(1)})\cr
Target CNN kernel size of the last CNN-layer; ignored if only one
CNN-layer is used (i.e. if \code{number_of_cnn_layers} is 1). Although CNN
kernel size is always an integer, this value can be non-integer,
potentially affecting the kernel-sizes of intermediate layers (which are
geometrically interpolated between \code{kernel_size_0} and \code{kernel_size_end}).}

\item{filters_0}{(\code{numeric(1)})\cr
Target filter number of the first CNN-layer. Although CNN filter number is
always an integer, this value can be non-integer, potentially affecting
the filter-numbers of intermediate layers (which are geometrically
interpolated between \code{filters_0} and \code{filters_end}).\cr
Note that filters are constant within convolutional blocks when
\code{residual_block} is \code{TRUE}.}

\item{filters_end}{(\code{numeric(1)})\cr
Target filter number of the last CNN-layer; ignored if only one CNN-layer
is used (i.e. if \code{number_of_cnn_layers} is 1). Although CNN filter number
is always an integer, this value can be non-integer, potentially affecting
the filter-numbers of intermediate dilation_rates layers (which are geometrically
interpolated between \code{kernel_size_0} and \code{kernel_size_end}).\cr
Note that filters are constant within convolutional blocks when
\code{residual_block} is \code{TRUE}.}

\item{dilation_end}{(\code{numeric(1)})\cr
Dilation of the last CNN-layer \emph{within each block}. Dilation rates within
each convolutional block grows exponentially from 1 (no dilation) for the
first CNN-layer to each block, to this value. Set to 1 (default) to
disable dilation.}

\item{max_pool_end}{(\code{numeric(1)})\cr
Target total effective pooling of CNN part of the network. "Effective
pooling" here is the product of the pooling rates of all previous
CNN-layers. A network with three CNN-layers, all of which are followed
by pooling layers of size 2, therefore has effective pooling of 8, with
the effective pooling at intermediate positions being 1 (beginning), 2,
and 4. Effective pooling after each layer is set to the power of 2 that is,
on a logarithmic scale, closest to
\verb{max_pool_end ^ (<CNN layer number> / <total number of CNN layers>)}.
Therefore, even though the total effective pooling size of the whole
CNN part of the network will always be a power of 2, having different,
possibly non-integer values of \code{max_pool_end}, will still lead to
different networks.}

\item{dense_layer_num}{(integer \code{numeric(1)})\cr
number of dense layers at the end of the network, not counting the output
layer.}

\item{dense_layer_units}{(integer \code{numeric(1)})\cr
Number of units in each dense layer, except for the output layer.}

\item{dropout_lstm}{Fraction of the units to drop for inputs.}

\item{dropout}{(\code{numeric(1)})\cr
Dropout rate of dense layers, except for the output layer.}

\item{batch_norm_momentum}{(\code{numeric(1)})\cr
\code{momentum}-parameter of \code{layer_batch_normalization} layers used in the
convolutional part of the network.}

\item{leaky_relu_alpha}{(\code{numeric(1)})\cr
\code{alpha}-parameter of the \code{layer_activation_leaky_relu} activation layers
used in the convolutional part of the network.}

\item{dense_activation}{(\code{character(1)})\cr
Which activation function to use for dense layers. Should be one of
\code{"relu"}, \code{"sigmoid"}, or \code{"tanh"}.}

\item{skip_block_fraction}{(\code{numeric(1)})\cr
What fraction of the first convolutional blocks to skip.
Only used when \code{model_type} is \code{"gap"}.}

\item{residual_block}{(\code{logical(1)})\cr
Whether to use residual layers in the convolutional part of the network.}

\item{reverse_encoding}{(\code{logical(1)})\cr
Whether the network should have a second input for reverse-complement
sequences.}

\item{optimizer}{(\code{character(1)})\cr
Which optimizer to use. One of \code{"adam"}, \code{"adagrad"}, \code{"rmsprop"}, or \code{"sgd"}.}

\item{model_type}{(\code{character(1)})\cr
Whether to use the global average pooling (\code{"gap"}) or recurrent
(\code{"recurrent"}) model type.}

\item{recurrent_type}{(\code{character(1)})\cr
Which recurrent network type to use. One of \code{"lstm"} or \code{"gru"}.
Only used when \code{model_type} is \code{"recurrent"}.}

\item{recurrent_layers}{(integer \code{numeric(1)})\cr
Number of recurrent layers.
Only used when \code{model_type} is \code{"recurrent"}.}

\item{recurrent_bidirectional}{(\code{logical(1)})\cr
Whether to use bidirectional recurrent layers.
Only used when \code{model_type} is \code{"recurrent"}.}

\item{recurrent_units}{(integer \code{numeric(1)})\cr
Number of units in each recurrent layer.
Only used when \code{model_type} is \code{"recurrent"}.}

\item{vocabulary_size}{(integer \code{numeric(1)})\cr
Vocabulary size of (one-hot encoded) input strings. This determines the
input tensor shape, together with \code{maxlen}.}

\item{last_layer_activation}{Either \code{"sigmoid"} or \code{"softmax"}.}

\item{loss_fn}{Either \code{"categorical_crossentropy"} or \code{"binary_crossentropy"}. If \code{label_noise_matrix} given, will use custom \code{"noisy_loss"}.}

\item{auc_metric}{Whether to add AUC metric.}

\item{num_targets}{(integer \code{numeric(1)})\cr
Number of output units to create.}

\item{model_seed}{Set seed for model parameters in tensorflow if not \code{NULL}.}

\item{bal_acc}{Whether to add balanced accuracy.}

\item{f1_metric}{Whether to add F1 metric.}

\item{mixed_precision}{Whether to use mixed precision (https://www.tensorflow.org/guide/mixed_precision).}

\item{mirrored_strategy}{Whether to use distributed mirrored strategy. If NULL, will use distributed mirrored strategy only if >1 GPU available.}
}
\value{
A keras model.

A keras model implementing genomenet architecture.
}
\description{
Create GenomeNet Model with Given Architecture Parameters
}
\examples{
\dontshow{if (reticulate::py_module_available("tensorflow")) (if (getRversion() >= "3.4") withAutoprint else force)(\{ # examplesIf}
model <- create_model_genomenet()
model
\dontshow{\}) # examplesIf}
}