Skip to main content

Table 4 Performance comparison of the distilled models trained with different combinations of losses

From: Improving the recall of biomedical named entity recognition with label re-correction and knowledge distillation

\({L}_{crf}\)

\({L}_{crf}^{T}\)

\({L}_{KD\_sim}^{T}\)

\({L}_{{l_{1} \_sim}}^{T}\)

\({L}_{{l_{2} \_sim}}^{T}\)

Adv

F (%)

✔

     

89.99

 

✔

    

90.13

 

✔

✔

   

90.16

 

✔

 

✔

  

90.13

 

✔

  

✔

 

90.35

 

✔

  

✔

✔

90.16

  1. The highest scores are highlighted in bold
  2. Adv: the short for adversarial learning
Back to article page

BMC Bioinformatics

ISSN: 1471-2105

Contact us