Difference between revisions of "Transformer"
| Line 6: | Line 6: | ||
*[[Natural Language Inference (NLI) and Recognizing Textual Entailment (RTE)]] | *[[Natural Language Inference (NLI) and Recognizing Textual Entailment (RTE)]] | ||
| − | “Attend” to specific parts of an image in sequence, one after another. By relying on a sequence of glances, they capture visual structure, can be contrasted with other machine vision techniques that process a whole image in a single, forward pass. | + | “Attend” to specific parts of the input (an image or text) in sequence, one after another. By relying on a sequence of glances, they capture (visual) structure, can be contrasted with other (machine vision) techniques that process a whole input e.g. image in a single, forward pass. |
<youtube>W2rWgXJBZhU</youtube> | <youtube>W2rWgXJBZhU</youtube> | ||
Revision as of 21:58, 10 May 2018
- Sequence to Sequence (Seq2Seq)
- Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM)
- Autoencoders / Encoder-Decoders
- Natural Language Inference (NLI) and Recognizing Textual Entailment (RTE)
“Attend” to specific parts of the input (an image or text) in sequence, one after another. By relying on a sequence of glances, they capture (visual) structure, can be contrasted with other (machine vision) techniques that process a whole input e.g. image in a single, forward pass.
