prevision_quantum_nn.models.utilities.early_stopper

early stopper module

Module Contents

Classes

EarlyStopper Class EarlyStopper.
class prevision_quantum_nn.models.utilities.early_stopper.EarlyStopper(window=10)

Class EarlyStopper.

Used to early stop the convergence of the quantum neural network

window

early stopper window or patience

Type:int
val_losses

list of validation losses to be kept in memory

Type:list
best_var

the best set of weights that yielded the minimum loss in the window

Type:list
best_val_loss

best validation loss so far in the window

Type:float
buffer

keeps the weights of the models until early stopper criterion is met

Type:queue
add_validation_loss(self, loss)

Adds validation loss to loss buffer.

Also stores the best loss reached so far.

Parameters:loss (float) – loss of the current iteration
get_stopping_criterion(self)

Stopping_criterions.

Returns:
bool
True is algorithm should stop False if algorithm should continue
Return type:stopping_criterion
update(self, val_loss, var)

Updates the early stopper at a given iteration.

Parameters:
  • val_loss (float) – validation loss to be stored
  • var – (list):list of weights that yielded this loss
Returns:

bool

if True, the calculatino will stop

Return type:

stopping_criterion

get_best_var(self)

Returns best weights.

Returns:
list
list of best weights
Return type:best_var