s.TFN: Neural Network with tensorflow [C, R]

Description

Train a Neural Network using keras and tensorflow

Usage

s.TFN(x, y = NULL, x.test = NULL, y.test = NULL, x.valid = NULL,
  y.valid = NULL, upsample = FALSE, upsample.seed = NULL,
  net = NULL, n.hidden.nodes = NULL,
  initializer = c("glorot_uniform", "glorot_normal", "he_uniform",
  "he_normal", "lecun_uniform", "lecun_normal", "random_uniform",
  "random_normal", "variance_scaling", "truncated_normal", "orthogonal",
  "zeros", "ones", "constant"), initializer.seed = NULL, dropout = 0,
  activation = c("relu", "selu", "elu", "sigmoid", "hard_sigmoid",
  "tanh", "exponential", "linear", "softmax", "softplus", "softsign"),
  l1 = 0, l2 = 0, batch.normalization = TRUE, output = NULL,
  loss = NULL, optimizer = c("rmsprop", "adadelta", "adagrad", "adam",
  "adamax", "nadam", "sgd"), learning.rate = NULL, metric = NULL,
  epochs = 50, batch.size = NULL, validation.split = 0.2,
  callback = keras::callback_early_stopping(patience = 150),
  scale = TRUE, x.name = NULL, y.name = NULL, print.plot = TRUE,
  print.error.plot = NULL, rtlayout.mat = c(2, 1),
  plot.fitted = NULL, plot.predicted = NULL,
  plot.theme = getOption("rt.fit.theme", "lightgrid"), question = NULL,
  verbose = TRUE, verbose.checkpoint = FALSE, outdir = NULL,
  save.mod = ifelse(!is.null(outdir), TRUE, FALSE), ...)

Arguments

Numeric vector or matrix / data frame of features i.e. independent variables

Numeric vector of outcome, i.e. dependent variable

x.test

Numeric vector or matrix / data frame of testing set features Columns must correspond to columns in x

y.test

Numeric vector of testing set outcome

upsample

Logical: If TRUE, upsample cases to balance outcome classes (for Classification only) Caution: upsample will randomly sample with replacement if the length of the majority class is more than double the length of the class you are upsampling, thereby introducing randomness

upsample.seed

Integer: If provided, will be used to set the seed during upsampling. Default = NULL (random seed)

n.hidden.nodes

Integer vector: Length must be equal to the number of hidden layers you wish to create. Can be zero (~GLM)

initializer

String: Initializer to use for each layer: "glorot_uniform", "glorot_normal", "he_uniform", "he_normal", "cun_uniform", "lecun_normal", "random_uniform", "random_normal", "variance_scaling", "truncated_normal", "orthogonal", "zeros", "ones", "constant". Glorot is also known as Xavier initialization. Default = "glorot_uniform"

initializer.seed

Integer: Seed to use for each initializer for reproducibility. Default = NULL

dropout

Floar, vector, (0, 1): Probability of dropping nodes. Can be a vector of length equal to N of layers, otherwise will be recycled. Default = 0

activation

String vector: Activation type to use: "relu", "selu", "elu", "sigmoid", "hard_sigmoid", "tanh", "exponential", "linear", "softmax", "softplus", "softsign". Default = "relu"

batch.normalization

Logical: If TRUE, batch normalize after each hidden layer. Default = TRUE

output

String: Activation to use for output layer. Can be any as in activation. Default = "linear" for Regression, "sigmoid" for binary classification, "softmax" for multiclass

loss

String: Loss to use: Default = "mean_squared_error" for regression, "binary_crossentropy" for binary classification, "sparse_categorical_crossentropy" for multiclass

optimizer

String: Optimization to use: "rmsprop", "adadelta", "adagrad", "adam", "adamax", "nadam", "sgd". Default = "rmsprop"

learning.rate

Float: learning rate. Defaults depend on optimizer used and are: rmsprop = .001, adadelta = 1, adagrad = .01, adamax = .002, adam = .001, nadam = .002, sgd = .1

metric

String: Metric used for evaluation during train. Default = "mse" for regression, "accuracy" for classification.

epochs

Integer: Number of epochs. Default = 100

batch.size

Integer: Batch size. Default = N of cases

validation.split

Float (0, 1): proportion of training data to use for validation. Default = .2

callback

Function to be called by keras during fitting. Default = keras::callback_early_stopping(patience = 150) for early stopping.

scale

Logical: If TRUE, scale featues before training. Default = TRUE column means and standard deviation will be saved in rtMod$extra field to allow scaling ahead of prediction on new data

x.name

Character: Name for feature set

y.name

Character: Name for outcome

print.plot

Logical: if TRUE, produce plot using mplot3 Takes precedence over plot.fitted and plot.predicted

plot.fitted

Logical: if TRUE, plot True (y) vs Fitted

plot.predicted

Logical: if TRUE, plot True (y.test) vs Predicted. Requires x.test and y.test

plot.theme

String: "zero", "dark", "box", "darkbox"

question

String: the question you are attempting to answer with this model, in plain language.

verbose

Logical: If TRUE, print summary to screen.

outdir

Path to output directory. If defined, will save Predicted vs. True plot, if available, as well as full model output, if save.mod is TRUE

save.mod

Logical. If TRUE, save all output as RDS file in outdir save.mod is TRUE by default if an outdir is defined. If set to TRUE, and no outdir is defined, outdir defaults to paste0("./s.", mod.name)

...

Additional parameters

Details

For more information on argument and hyperparameters, see (https://keras.rstudio.com/) and (https://keras.io/) It is important to define network structure and adjust hyperparameters. You cannot expect defaults to work on any dataset.

Description

Usage

Arguments

Details

See Also