search

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
681 stars 217 forks source link

data preparation #10

Closed sue2xlh closed 3 years ago

sue2xlh commented 3 years ago

HI, thank you for your sharing, read you codes i have some questuons. during feature transform, will need to normalize the dense features? Deep networks are said to be sensitive to dense features. and sparse feature, why use one-hot encoder instead of label encoder?

alvin319 commented 3 years ago

Feature normalization is highly dependent on the dataset and for the Census Income example, I just stuck with the one-hot encoding for brevity and the paper didn't really clarify their feature engineering approach.