Module compute::optimize[][src]

Expand description

Various optimization algorithms (eg. Adam, SGD, Levenberg-Marquardt).

Structs

Implements the Adam optimizer. See Kingma and Ba 2014 for details about the algorithm.

Implements a Levenberg-Marquardt optimizer for solving (non-linear) least squares problems.

Implements the Stochastic Gradient Descent optimizer with (Nesterov) momentum.

Tape (Wengert list) that tracks differentiable variables, intermediate values, and the operations applied to each.

Differentiable variable. This is the main type that users will interact with.

Traits

Calculate gradients with respect to particular variables.

Trait for calculating expressions and tracking gradients for float power operations.