ZGen::BestFirst::Learner Class Reference

#include <learner.h>

Public Types
enum	{ CONVERGED = 0, SAFE_UNCONVERGED, UNSAFE_UNCONVERGED, NO_GOLD_IN_AGENDA }
	The learning result status. More...

Public Member Functions
	Learner (Model *model, int max_iteration=1)
void	config_model (Model *model)
void	config_timestamp (int timestamp)
int	learn (const SpanSet &gold_spans, SpanMaxHeap &agenda)

Member Enumeration Documentation

anonymous enum

The learning result status.

Enumerator
CONVERGED
SAFE_UNCONVERGED
UNSAFE_UNCONVERGED
NO_GOLD_IN_AGENDA

Constructor & Destructor Documentation

ZGen::BestFirst::Learner::Learner	(	Model *	model,
		int	max_iteration = 1
	)

Member Function Documentation

void ZGen::BestFirst::Learner::config_model

(

Model *

model

)

Config the model onto the learner.

Parameters

[in]

model

The pointer to the model.

void ZGen::BestFirst::Learner::config_timestamp

(

int

timestamp

)

Config the timestamp to learner.

Parameters

[in]

timestamp

The timestamp.

int ZGen::BestFirst::Learner::learn	(	const SpanSet &	gold_spans,
		SpanMaxHeap &	agenda
	)

The main function for Perceptron learning. The learning process includes:

• 1: calculate the number of gold spans in the agenda.
• 2: check whether the current agenda is converged.
• 3: while not converged: update parameter, check converged

PREREQUISITE

• model is configed.
• timestamp is configed.

Parameters

[in]	gold_spans	The set of gold spans.
	in/out]	agenda The agenda.

Returns

int The learning status.

---

The documentation for this class was generated from the following files:

• src/bestfirst/learning/learner.h
• src/bestfirst/learning/learner.cc