
1. The total complexity per layer is reduced
2. The amount of computation that can be parallelized, as measured by the minimum number of sequential operations required also reduces.
3. Reduction of the path length between long-range dependencies in the network.


New York University

1. This model generally has an encoder - decoder structure.
2. The encoder maps an input sequence of symbol representations to a sequence of continuous representations.
3. The decoder then generates an output sequence on symbols, one element at a time.


Learn Before

Related