AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Geometric Deep Learning and Causal Model Learning
There's a kind of analogy between geometric deep learning and causal representation learning. If you have a causal model, ah, then if you perturb the input, the prediction that you get out of it remains a valid, a valid output. Where is, if it's non causal, it has te potential to learn all these kind of spurious, o spurious structures. And i wonder where the neuro networks can find abstractions. Would love to get your comment on that.