BriefGPT.xyz
Nov, 2018
领域自适应迁移学习的专家模型
Domain Adaptive Transfer Learning with Specialist Models
HTML
PDF
Jiquan Ngiam, Daiyi Peng, Vijay Vasudevan, Simon Kornblith, Quoc V. Le...
TL;DR
本文研究了传递学习的有效性，并发现原先认为的更多的预训练数据不一定就能提高模型性能，而是需要审慎选择预训练数据。同时，作者提出了基于目标数据集计算重要性权重的领域自适应传递学习方法，并在多个细粒度分类数据集上取得了最先进的结果。
Abstract
transfer learning
is a widely used method to build high performing
computer vision
models. In this paper, we study the efficacy of
transfer learn
→