seli.opt

Contents

seli.opt#

Optimizers and loss functions for training neural networks.

Functions#

get_arrays(module[, collection])

Extract arrays from parameters in a module.

grad(func, *[, collection, has_aux])

Create a function that computes gradients with respect to module parameters.

set_arrays(module, arrays)

Set arrays back into parameters in a module.

value_and_grad(func, *[, collection, has_aux])

Create a function that computes gradients with respect to module parameters.

Classes#

Adagrad([lr, eps])

Adaptive Gradient optimizer.

Adam([lr, beta1, beta2, eps])

Adaptive Moment Estimation optimizer.

BinaryCrossEntropy()

Binary cross entropy loss function.

Loss()

Base class for all loss functions.

MeanAbsoluteError()

Mean absolute error loss function.

MeanSquaredError()

Mean squared error loss function.

Momentum([lr, beta])

Momentum optimizer.

Nesterov([lr, beta])

Nesterov Accelerated Gradient optimizer.

Optimizer()

Base class for all gradient basedoptimizers.

RMSProp([lr, beta, eps])

Root Mean Square Propagation optimizer.

SGD([lr])

Stochastic Gradient Descent optimizer.
Contents