What optimization algorithms can be used for training with the crfTrain action?
3 vues
Réponse
You can specify the optimization solver using the 'algorithm' subparameter within 'nloOpts'. The available options are: 'LBFGS' (limited-memory Broyden-Fletcher-Goldfarb-Shanno), 'SGD' (stochastic gradient descent), 'ADAM' (adaptive moments variant of SGD), and 'HF' (Hessian free, which is experimental).
