arXiv:1706.03762 · cs.CL · 2017
Attention Is All You Need
Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, Illia Polosukhin
피인용 191,347회 · 영향력 있는 인용 20,648회— Semantic Scholar
TL;DR · Semantic Scholar
A new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely is proposed, which generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data.
한국어 요약
AewolFlow 독자가 읽으며 생성한 AI 요약입니다 — 논문 본문에 근거하며, 정확한 내용은 원문으로 확인하세요.
문제: 기존 RNN 기반 모델은 순차적 연산으로 인해 병렬 처리가 불가능했습니다. CNN 역시 멀리 떨어진 단어 간의 의존성을 학습하는 데 어려움이 있었습니다.
방법: recurrence와 convolution을 배제하고 오직 self-attention에만 의존하는 Transformer 아키텍처를 제안합니다. 효율적인 정보 처리를 위해 Scaled Dot-Product Attention과 Multi-Head Attention을 도입했습니다.
결과: WMT 2014 English-to-German 태스크에서 28.4 BLEU를 기록하며 기존 성능을 능가했습니다. 또한 학습 속도가 크게 향상되었고 English constituency parsing에서도 우수한 성능을 증명했습니다.
한계: 긴 시퀀스를 처리할 때 self-attention의 계산량이 증가하므로 restricted self-attention에 대한 추가 연구가 요구됩니다. 또한 이미지나 오디오 등 다양한 modality로의 확장이 과제로 남아있습니다.
Abstract
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.
출처: https://arxiv.org/abs/1706.03762 — Thank you to arXiv for use of its open access interoperability.
원문·번역·핵심 문장·AI 질문까지 — 이 논문을 AewolFlow에서 이어 읽으세요.