PWL #58: Attention Is All You Need

Every 1st Thursday of the month

Location image of event venue

Details

Details
• What we'll do
Lightning Talk: Marvin Humphrey, visiting from the San Diego chapter of Papers We Love, will present on the evolution of error handling. http://joeduffyblog.com/2016/02/07/the-error-model/

Main Event: Alec Morgan will be presenting "Attention Is All You Need" (https://arxiv.org/abs/1706.03762)
"The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data."
(Talking to Alec last month, it's a very important modern paper for Natural Language Processing)

• Important to know

Big ups to Glympse for hosting this month!

As a chapter of Papers We Love we abide by and enforce the PWL Code of Conduct (https://github.com/papers-we-love/seattle/blob/master/code-of-conduct.md) at our events. Please give it a read, plan on acting like an adult, and involve one of the organizers if you need help.

Stop slacking and join us in the #seattle channel at https://papersweloveslack.herokuapp.com!

If you have a paper you'd like to present, or even just a mini, please hit up one of the organizers :) We're always looking for more presenters.