Module crf 

Linear-chain Conditional Random Fields (CRF).

Re-exports

pub use crf_train::LbfgsConfig;

pub use crf_train::crf_log_likelihood_and_gradient;

pub use crf_train::train_crf_lbfgs;

pub use general_graph::Edge;

pub use general_graph::GeneralGraphCrf;

pub use general_graph::GraphCrfConfig;

pub use lbfgs_b::LbfgsB;

pub use lbfgs_b::LbfgsBConfig;

pub use lbfgs_b::LbfgsBResult;

pub use linear_chain_crf::LinearChainCrf;

pub use neural_crf::NeuralCrf;

pub use neural_crf::NeuralCrfForward;

pub use neural_crf::NeuralCrfGrad;

pub use sgd::CrfSgd;

pub use sgd::CrfSgdConfig;

pub use skip_chain::SkipChainConfig;

pub use skip_chain::SkipChainCrf;

pub use viterbi_decode::viterbi_decode;

Modules

crf_train

CRF training: log-likelihood + gradient via forward-backward in score space, plus a limited-memory BFGS (L-BFGS) optimiser with backtracking line search.

general_graph

General-graph CRF training via loopy belief propagation.

lbfgs_b

L-BFGS-B (Byrd, Lu, Nocedal 1995): limited-memory BFGS with box constraints lower ≤ x ≤ upper.

linear_chain_crf

Linear-chain CRF parameterisation.

neural_crf

Neural linear-chain Conditional Random Field.

sgd

CRF training via SGD / AdaGrad.

skip_chain

Skip-chain Conditional Random Fields.

viterbi_decode

Viterbi decoding for linear-chain CRFs (score-space).