d_H2OGLRM: Generalized Low-Rank Models (GLRM) on H2O

In egenn/rtemis: Machine Learning and Visualization

Generalized Low-Rank Models (GLRM) on H2O

Description

Perform GLRM decomposition using h2o::h2o.glrm Given Input matrix A: ⁠A(m x n) = X(m x k) \%*\% Y(k x n)⁠

Usage

d_H2OGLRM(
  x,
  x.test = NULL,
  x.valid = NULL,
  k = 3,
  ip = "localhost",
  port = 54321,
  transform = "NONE",
  loss = "Quadratic",
  regularization.x = "None",
  regularization.y = "None",
  gamma.x = 0,
  gamma.y = 0,
  max_iterations = 1000,
  max_updates = 2 * max_iterations,
  init_step_size = 1,
  min_step_size = 1e-04,
  seed = -1,
  init = "PlusPlus",
  svd.method = "Randomized",
  verbose = TRUE,
  print.plot = TRUE,
  plot.theme = rtTheme,
  n.cores = rtCores,
  ...
)

Arguments

x	Input data
x.test	Optional test set. Will be projected on to NMF basis
x.valid	Optional validation set
k	Integer: Rank of decomposition
ip	Character: IP address of H2O server. Default = "localhost"
port	Integer: Port number for server. Default = 54321
transform	Character: Transformation of input prior to decomposition
loss	Character: Numeric loss function: "Quadratic", "Absolute", "Huber", "Poisson", "Hinge", "Logistic", "Periodic". Default = "Quadratic"
regularization.x	Character: Regularization function for X matrix: "None", "Quadratic", "L2", "L1", "NonNegative", "OneSparse", "UnitOneSparse", "Simplex". Default = "None"
regularization.y	Character: Regularization function for Y matrix: "None", "Quadratic", "L2", "L1", "NonNegative", "OneSparse", "UnitOneSparse", "Simplex". Default = "None"
gamma.x	Float: Regularization weight on X matrix. Default = 0
gamma.y	Float: Regularization weight on Y matrix. Default = 0
max_iterations	Integer: Maximum number of iterations. Default = 1000
max_updates	Integer: Maximum number of iterations. Default = 2 * max_iterations
init_step_size	Float: Initial step size. Default = 1
min_step_size	Float: Minimum step size. Default = .0001
seed	Integer: Seed for random number generator. Default = -1 (time-based)
init	Character: Initialization mode: "Random", "SVD", "PlusPlus", "User". Default = "PlusPlus"
svd.method	Character: SVD method for initialization: "GramSVD", "Power", "Randomized". Default = "Randomized"
verbose	Logical: If TRUE, print console messages
print.plot	Logical: If TRUE, print objective score against iteration number
plot.theme	Character: Theme to pass to mplot3_xy if print.plot = TRUE
n.cores	Integer: Number of cores to use
...	Additional parameters to be passed to h2o::h2o.glrm

Details

Learn more about GLRM from the H2O tutorial https://github.com/h2oai/h2o-tutorials/blob/master/tutorials/glrm/glrm-tutorial.md

Value

rtDecom object

Author(s)

E.D. Gennatas

See Also

Other Decomposition: d_H2OAE(), d_ICA(), d_Isomap(), d_KPCA(), d_LLE(), d_MDS(), d_NMF(), d_PCA(), d_SPCA(), d_SVD(), d_TSNE(), d_UMAP()

egenn/rtemis documentation built on Oct. 28, 2024, 6:30 a.m.