Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
anyBodyCanDoDL
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
2 ms
·
1.
▲
by
anyBodyCanDoDL
8y ago
That's very ignorant to say that attention or convolution is anywhere close to the expressivity of RNNs(LSTM). Instead of reading a few cherry-picked results from a model with extremely tuned hyperparameters, pick any random set of tas