Hyperparameter Optimizers#
- jetgp.hyperparameter_optimizers.cobyla.cobyla(func, lb, ub, **kwargs)[source]#
COBYLA optimizer with bounds as inequality constraints and multi-start.
- Parameters:
func (callable) – Function to minimize.
lb (array-like) – Lower and upper bounds.
ub (array-like) – Lower and upper bounds.
kwargs (dict) – Optional arguments: - x0 : initial guess for first restart - n_restart_optimizer : number of random restarts (default=10) - debug : bool, print intermediate results (default=False) - Any COBYLA options: maxiter, rhobeg, catol, f_target
- jetgp.hyperparameter_optimizers.jade.jade(func, lb, ub, **kwargs)[source]#
Wrapper for JADE with unified interface.
- jetgp.hyperparameter_optimizers.lbfgs.lbfgs(func, lb, ub, **kwargs)[source]#
L-BFGS-B optimizer with intelligent restart strategies.
- Parameters:
func (callable) – Function to minimize.
lb (array-like) – Lower and upper bounds.
ub (array-like) – Lower and upper bounds.
kwargs (dict) –
- strategy’random’, ‘lhs’, ‘sobol’, ‘exclusion’, ‘adaptive’, ‘clustering’
(default=’adaptive’)
n_restart_optimizer : number of restarts (default=10)
x0 : initial guess for first restart
maxiter, ftol, gtol, debug, disp : L-BFGS-B options
descriptions (Strategy) –
‘random’: Pure random restarts (original behavior)
’lhs’: Latin Hypercube Sampling for space-filling coverage
’sobol’: Sobol sequence for low-discrepancy coverage
’exclusion’: Avoid regions near previously found optima
’adaptive’: Two-phase exploration/exploitation with basin estimation
’clustering’: Online clustering to identify and avoid basins
- jetgp.hyperparameter_optimizers.lbfgs.lbfgs_smart(func, lb, ub, **kwargs)[source]#
Smart L-BFGS-B optimizer with intelligent restart strategies.
Strategies: - ‘lhs’: Latin Hypercube Sampling for space-filling initial points - ‘sobol’: Sobol quasi-random sequence for low-discrepancy coverage - ‘exclusion’: Avoid regions near previous optima - ‘adaptive’: Combine exclusion with basin size estimation
- Parameters:
func (callable) – Function to minimize.
lb (array-like) – Lower and upper bounds.
ub (array-like) – Lower and upper bounds.
kwargs (dict) –
x0 : initial guess for first restart
n_restart_optimizer : number of restarts (default=10)
strategy : ‘random’, ‘lhs’, ‘sobol’, ‘exclusion’, ‘adaptive’ (default=’exclusion’)
exclusion_radius : fraction of domain to exclude around optima (default=0.1)
max_rejection : max attempts to find valid starting point (default=100)
maxiter, ftol, gtol, debug, disp : standard L-BFGS-B options
- jetgp.hyperparameter_optimizers.powell.powell(func, lb, ub, **kwargs)[source]#
Powell’s method with bounds and multi-start.
- Parameters:
func (callable) – Function to minimize.
lb (array-like) – Lower and upper bounds.
ub (array-like) – Lower and upper bounds.
kwargs (dict) – Optional arguments: - x0 : initial guess for first restart - n_restart_optimizer : number of random restarts (default=10) - debug : bool, print intermediate results (default=False) - Any Powell-specific options: maxiter, xtol, ftol, disp