some questions about model effect

drawbridge / keras-mmoe

A TensorFlow Keras implementation of "Modeling Task Relationships in Multi-task Learning with Multi-gate Mixture-of-Experts" (KDD 2018)

MIT License

690 stars 219 forks source link

some questions about model effect #12

Closed H-H-J closed 3 years ago

H-H-J commented 3 years ago

Hi, thanks for your share. Have you ever encountered the following problem？When the average predictive value of one task is 0.74, the value of the other task is as high as 0.92. However, the proportion of positive samples of this task is only about 12.5%. Whether the label column of the previous task (the predicted average value is 0.74) affected the second task?

alvin319 commented 3 years ago

Hi! I don't know if I can answer that question given this observation that the modeling effect is highly correlated with the dataset itself.