ParentFull threadinciampati·One 3-layer (1 hidden layer) neural network can already approximate anything. You don't even need to stack them.View on HN