empulse.optimizers

The optimizers module contains the optimizers used to optimize models in the empulse.models module.

Optimizers

Optimizer	Abstract base class for all logit model optimizers.
Adam	Adam optimizer with optional AMSGrad correction.
GeneticAlgorithmOptimizer	Real-coded Genetic Algorithm (RGA) optimizer for logit models.
LBFGSBOptimizer	Limited-memory BFGS with box constraints (L-BFGS-B) via scipy.optimize.minimize.
MemeticOptimizer	Real-coded Lamarckian Memetic Algorithm optimizer for logit models.
RMSProp	RMSProp optimizer.
ScipyOptimizer	General-purpose wrapper around scipy.optimize.minimize.
SGD	Stochastic Gradient Descent with optional Nesterov momentum.

Optimizer Components

Used to construct optimizers in the empulse.optimizers module.

Generation	A single generation of a Real-coded Genetic Algorithm (RGA).
LamarckianGeneration	Real-coded GA generation with Lamarckian local gradient search.

Schedulers

Used to schedule the learning rate or smoothing parameters during training.

Schedule	Abstract base class for parameter schedules.
ConstantSchedule	Constant schedule – always returns value.
CosineAnnealingSchedule	Cosine annealing between max_value and min_value over t_max epochs.
ExponentialSchedule	Exponential schedule: value = start_value * gamma ** epoch.
LinearSchedule	Linear interpolation from start_value to end_value over n_steps epochs.
StepSchedule	Step decay: multiply by gamma every step_size epochs.
WarmupSchedule	Linear warm-up for warmup_steps epochs, then delegates to after_schedule.