AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
The Importance of Meta-Learning a Loss Function
Since then have you seen anyone use this idea of meta-learning a loss function in other contexts? Yeah so people have actually used it in a couple of different scenarios and some of them I don't even know if they kind of realized that use it in the the imitation learning work. Another place that people have used it is if it tests if when you're trying to learn and you test you have both a label training set and an unlabeled test set then you might want to use a learned loss function for those situations.