Find the next beta according to: ¤ approximation_ratio(alpha, beta_old, beta_new) == max_approx ¤ beta_new >= beta_old + min_increase ¤ beta_new <= beta_max — next_beta • slise

Find the next beta according to: ¤ approximation_ratio(alpha, beta_old, beta_new) == max_approx ¤ beta_new >= beta_old + min_increase ¤ beta_new <= beta_max

Usage

next_beta(
  residuals2,
  epsilon2,
  beta = 0,
  weight = NULL,
  beta_max = 20/epsilon2,
  log_max_approx = log(1.15),
  beta_min_increase = (beta_max + beta) * 5e-04
)

Arguments

residuals2: squared residuals
epsilon2: squared error tolerance
beta: current sigmoid steepness
weight: weight vector (default: NULL)
beta_max: max sigmoid steepnsess
log_max_approx: logarithm of the approximation ratio target for increasing beta
beta_min_increase: minimum beta step

Value

beta_new