Reranker Framework (ReFr)
Reranking framework for structure prediction and discriminative language modeling
|
An inner interface specifying an update function for a model. More...
#include <model.H>
Public Member Functions | |
virtual void | Update (Model *model, CandidateSet &example)=0 |
Updates this model based on the specified training example. More... | |
Public Member Functions inherited from reranker::FactoryConstructible | |
virtual | ~FactoryConstructible () |
virtual void | RegisterInitializers (Initializers &initializers) |
virtual void | Init (const Environment *env, const string &arg) |
An inner interface specifying an update function for a model.
The update function typically goes hand in hand with the type of model. For example, perceptron-style models get updated by adding to the current weight vector a possibly scaled version of the “gold” candidate hypothesis and subtracting a possibly scaled version of the candidate hypothesis that scored highest under the current model. Typically, implementations of this interface will be inner classes of the concrete Model implementation to which they apply.
|
pure virtual |
Updates this model based on the specified training example.
model | the model to update |
example | the current training example that caused the model to need updating |