Ensemble Learning

[flassTer]

Hello NVIDIA,

I am training some models on different datasets. It has come to my notice that when transfer learning, the model starts becoming better for the new dataset but its performance worsens on the previous. Therefore I decided to use ensemble learning to fix this. I have considered using AdaBoost (https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.AdaBoostClassifier.html)

How could the sklearn implementation of this method be used with OpenSeq2Seq or an ensemble learning boosting approach in general?

[borisgin]

Have you looked at ScipyOptimizerInterface API? You can also implement new optimizer in TF (see for example tf/contrib/optimizers