dnnet.backprop.r: Back Propagation

In SkadiEye/deepTL: Deep Treatment Learning

Back Propagation

Description

Back Propagation

Usage

dnnet.backprop.r(
  n.hidden,
  w.ini,
  load.param,
  initial.param,
  x,
  y,
  w,
  valid,
  x.valid,
  y.valid,
  w.valid,
  activate,
  activate_,
  n.epoch,
  n.batch,
  model.type,
  learning.rate,
  l1.reg,
  l2.reg,
  early.stop,
  early.stop.det,
  learning.rate.adaptive,
  rho,
  epsilon,
  beta1,
  beta2,
  loss.f
)

Arguments

n.hidden	A numeric vector for numbers of nodes for all hidden layers.
w.ini	Initial weight parameter.
load.param	Whether initial parameters are loaded into the model.
initial.param	The initial parameters to be loaded.
x	x
y	y
w	w
valid	If exists the validation set
x.valid	x-valid
y.valid	y-valid
w.valid	w-valid
activate	Activation Function.
activate_	The forst derivative of the activation function.
n.epoch	Maximum number of epochs.
n.batch	Batch size for batch gradient descent.
model.type	Type of model.
learning.rate	Initial learning rate, 0.001 by default; If "adam" is chosen as an adaptive learning rate adjustment method, 0.1 by defalut.
l1.reg	weight for l1 regularization, optional.
l2.reg	weight for l2 regularization, optional.
early.stop	Indicate whether early stop is used (only if there exists a validation set).
early.stop.det	Number of epochs of increasing loss to determine the early stop.
learning.rate.adaptive	Adaptive learning rate adjustment methods, one of the following, "constant", "adadelta", "adagrad", "momentum", "adam".
rho	A parameter used in momentum.
epsilon	A parameter used in Adagrad and Adam.
beta1	A parameter used in Adam.
beta2	A parameter used in Adam.
loss.f	Loss function of choice.

Value

Returns a list of results to dnnet.

SkadiEye/deepTL documentation built on Nov. 17, 2022, 1:41 p.m.