A Comparison of Update Strategies for Large-Scale Maximum Expected BLEU Training
23046Knowledge Triples
| Source Entity | Relationship | Target Entity |
|---|---|---|
|
Discriminative Training
Method |
performedOn |
BOLT Chinese-English
Task |
|
Discriminative Training
Method |
performedOn |
WMT German-English
Task |
|
Discriminative Training
Method |
uses |
RPROP
Method |
|
RPROP
Method |
comparesWith |
ADAGRAD
Method |
|
RPROP
Method |
comparesWith |
Growth Transformation
Method |
|
RPROP
Method |
comparesWith |
Stochastic Gradient Descent
Method |
|
RPROP
Method |
optimizes |
Maximum Expected BLEU
Metric |
|
WMT German-English
Task |
trainedOn |
4M Sentence Pairs
Material |