Closed ethanjperez closed 6 years ago
Actually we do mention LSTMs, but it's more of an aside in the Image Recognition section. Should we move this to the NLP section and/or highlight it more?
Yes, I agree that it is more natural to start the NLP section with an LSTM reference than with the Gated-Attention Reader. Do you want to give it a try?
Is there a reason we didn't mention that LSTM's use feature-wise transformations? I feel like that's one of the biggest wins of feature-wise transformations. Maybe we don't need a diagram for it, but should we at least mention it somewhere (i.e. in the NLP section)?