1,295 followers
@Yoni_Sch This can be done with adversarial training, in the same way we use it to explicitly promote domain invariance (https://t.co/TWUW6owkcP). We can force the representation to be orthogonal to known predictors.
@Yoni_Sch This can be done with adversarial training, in the same way we use it to explicitly promote domain invariance (https://t.co/TWUW6owkcP). We can force the representation to be orthogonal to known predictors.
@MoritzGerstung @jnirsch @Steffen_Vogler_ @MAubreville @anantm Exactly, we did this using domain adversarial learning (but in principle the same method can be used with any confounding variable): https://t.co/TWUW6owkcP
Learning Domain-Invariant Representations of Histological Images. https://t.co/MDCeFRo8ud