One Model To Learn Them All
Łukasz Kaiser Google Brain、Aidan N. Gomez University of Toronto、Noam Shazeer Google Brain、Ashish Vaswani Google Brain、Niki Parmar Google Research、Llion Jones Google Research、Jakob Uszkoreit Google Research Work performed while at Google Brain.
Abstract
Deep learning yields great results across many fields, from speech recognition, image classification, to translation. But for each problem, getting a deep model to work well involves research into the architecture and a long period of tuning. We present a single model that yields good results on a number of problems spanning multiple domains. In particular, this single model is trained concurrently on ImageNet, multiple translation tasks, image captioning (COCO dataset), a speech recognition corpus, and an English parsing task. Our model architecture incorporates building blocks from multiple domains. It contains convolutional layers, an attention mechanism, and sparsely-gated layers. Each of these computational blocks is crucial for a subset of the tasks we train on. Interestingly, even if a block is not crucial for a task, we observe that adding it never hurts performance and in most cases improves it on all tasks. We also show that tasks with less data benefit largely from joint training with other tasks, while performance on large tasks degrades only slightly if at all. ††Code available at https://github.com/tensorflow/tensor2tensor
中文速览
深度学习在语音、视觉、翻译等各领域都表现出色，但每换一个任务就得重新设计网络结构并反复调参，既耗时又难以复用。谷歌研究者提出了一个名为 MultiModel 的统一架构，用一套模型同时在图像分类（ImageNet）、图像描述（COCO）、多语言翻译（英德、英法互译）、语音识别和英语句法分析共八个任务上联合训练，核心思路是：用极轻量的"模态子网络（modality net）"把图像、音频、文本等不同形式的输入统一映射到同一表示空间，再由包含深度可分离卷积、注意力机制和稀疏门控混合专家（sparsely-gated mixture-of-experts）三类计算模块的共享主体进行处理。实验结果表明，单个模型在所有任务上都达到了不错的水准，数据量少的任务从联合训练中获益显著，而数据量大的任务性能几乎没有损失；更有趣的是，即便某个计算模块对某类任务并非必需，加上它也不会拖累性能，反而往往略有提升。这项工作首次证明了跨模态、跨领域的多任务统一深度学习模型是可行的，为减少任务专属工程代价、迈向更通用的 AI 系统迈出了重要一步。
原文 arXiv:1706.05137;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1706.05137v1