set.control.coord.sgd: Check and set the control parameters for the coordinate-SGD...

In sgdGMF: Estimation of Generalized Matrix Factorization Models via Stochastic Gradient Descent

Check and set the control parameters for the coordinate-SGD algorithm

Description

Check if the input control parameters are allowed and set them to default values if they are not. Returns a list of well-defined control parameters.

Usage

set.control.coord.sgd(
  normalize = TRUE,
  maxiter = 1000,
  eps = 1e-08,
  nafill = 10,
  tol = 1e-08,
  size = c(100, 100),
  burn = 1,
  rate0 = 0.01,
  decay = 0.01,
  damping = 0.001,
  rate1 = 0.1,
  rate2 = 0.01,
  verbose = FALSE,
  frequency = 250,
  progress = FALSE
)

Arguments

normalize	if TRUE, normalize U and V to uncorrelated Gaussian U and upper triangular V with positive diagonal
maxiter	maximum number of iterations
eps	how much shrinkage has to be introduced on extreme predictions lying outside of the data range
nafill	how frequently the NA values are filled, by default NA values are filled at each iteration of the algorithm
tol	tolerance threshold for the stopping criterion
size	mini-batch size, the first value is for row sub-sample, the second value is for column sub-sample
burn	percentage of iterations to ignore before performing Polyak averaging
rate0	initial learning rate
decay	learning rate decay
damping	regularization parameter which is added to the Hessian to ensure numerical stability
rate1	exponential decay rate for the moment estimate of the gradient
rate2	exponential decay rate for the moment estimate of the Hessian
verbose	if TRUE, print the optimization status
frequency	how often the optimization status is printed (only if verbose=TRUE)
progress	if TRUE, print a compact progress-bar instead of a full-report of the optimization status (only if verbose=TRUE)

Value

A list of control parameters for the adaptive SGD algorithm with coordinate-wise sub-sampling

Examples

library(sgdGMF)

# Empty call
set.control.coord.sgd()

# Parametrized call
set.control.coord.sgd(maxiter = 2000, rate0 = 0.01, decay = 0.01)

sgdGMF documentation built on April 3, 2025, 7:37 p.m.