Create gradient-based optimization configuration — mle_config

Extends base configuration with gradient-specific parameters like learning rate and distance metric.

Usage

mle_config_gradient(
  eta = 1,
  norm = function(x) max(abs(x)),
  max_iter = 100L,
  abs_tol = NULL,
  rel_tol = 1e-05,
  trace = FALSE,
  debug = FALSE,
  debug_freq = 1L
)

Arguments

eta: Learning rate / step size (numeric, default: 1.0)
norm: Distance measure function (default: max absolute value)
max_iter: Maximum iterations (integer, default: 100)
abs_tol: Absolute tolerance (numeric or NULL, default: NULL to use rel_tol)
rel_tol: Relative tolerance (numeric, default: 1e-5)
trace: Store optimization path (logical, default: FALSE)
debug: Print debug information (logical, default: FALSE)
debug_freq: Debug output frequency (integer, default: 1)

Value

An mle_config_gradient object

Examples

# Basic gradient configuration
config <- mle_config_gradient(eta = 0.1, max_iter = 500)

# With custom norm (L2 norm)
config <- mle_config_gradient(
  eta = 0.01,
  norm = function(x) sqrt(sum(x^2))
)