ConceptAI & ML
Position-wise feed-forward networks
Position-wise feed-forward networks apply a simple multi-layer perceptron consisting of linear transformations and non-linearities independently to every token position.
Know first
Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.
Linear layer1ReLU non-linearity1
Lecturers
1Connections
2The lectures that define it or assume it, and the concepts next to it.