Struct BFGS 

pub struct BFGS { /* private fields */ }

Broyden-Fletcher-Goldfarb-Shanno optimization algorithm

This struct configures the BFGS quasi-Newton method, which approximates the inverse Hessian using rank-2 updates. It is as memory-intensive as regular Newton method (O(n^2) storage) but it does not require double differentiation.

Fields

• max_steps - Maximum number of optimization steps.
• gtol - Optional tolerance for gradient norm (stop if ||grad|| < gtol).
• ftol - Optional tolerance for change in objective value (stop if |f - prev_f| < ftol).

Implementations

impl BFGS

pub fn new(max_steps: usize, gtol: Option<f64>, ftol: Option<f64>) -> Self

Creates a new BFGS optimizer with the given parameters.

Arguments

• max_steps - Maximum iterations.
• gtol - Optional gradient tolerance.
• ftol - Optional function value change tolerance.

Returns

Configured BFGS instance.

Trait Implementations

impl Display for BFGS

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Optimizer for BFGS

fn optimize( &self, function: &dyn Fn(&Tensor) -> Tensor, x0: &Tensor, ) -> Result<Tensor>

Solves the problem of optimization of function function starting from point x0