fsbatch_control: Options for Fellner-Schall algorithm

In davidruegamer/deepregression: Fitting Deep Distributional Regression

Options for Fellner-Schall algorithm

Description

Options for Fellner-Schall algorithm

Usage

fsbatch_control(
  factor = 0.01,
  lr_scheduler = NULL,
  avg_over_past = FALSE,
  constantdiv = 0,
  constantinv = 0,
  constinv_scheduler = NULL
)

Arguments

factor	a factor defining how much of the past is used for the Fellner-Schall algorithm; defaults to 0.01.
lr_scheduler	a scheduler adapting factor in each step; defaults to NULL.
avg_over_past	logical, whether the beta coefficients should be averaged over the past values to stabilize estimation; defaults to FALSE
constantdiv	small positive constant to stabilize training in small batch regimes; defaults to 0.0.
constantinv	small positive constant to stabilize training in small batch regimes; defaults to 0.0.
constinv_scheduler	scheduler for constantinv; per default NULL which results in an exponential decay with rate 1

Value

Returns a list with options

davidruegamer/deepregression documentation built on May 30, 2022, 6:21 p.m.