search

attilanagy234 / neural-punctuator

Complimentary code for our paper Automatic punctuation restoration with BERT models
MIT License
48 stars 7 forks source link

Validation accuracy more than train #8

Closed UsamaI000 closed 3 years ago

UsamaI000 commented 3 years ago

Hi, I am getting validation accuracy greater than training and validation loss lesser than train. Can you tell why? Maybe due to class imbalance? I am using different dataset

bana513 commented 3 years ago

Hi, validation loss can be lower than train, e.g. using dropout during training. Best practice is to the model with early stopping on validation loss. We cannot say anything about your accuracy, especially if you are using different dataset. However, we recommend using F-score instead of accuracy with imbalanced datasets.