Optimizing Trainable Parameters

Optimizing Trainable Parameters

To add support for a new type of gradient-based parameter update, create a new type with the following methods defined for the types of generative functions that are to be supported.

Gen.init_update_state โ€” Function.
state = init_update_state(conf, gen_fn::GenerativeFunction, param_list::Vector)

Get the initial state for a parameter update to the given parameters of the given generative function.

param_list is a vector of references to parameters of gen_fn. conf configures the update.

source
Gen.apply_update! โ€” Function.
apply_update!(state)

Apply one parameter update, mutating the values of the trainable parameters, and possibly also the given state.

source