arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2009.14737cs.LGcs.CVstat.ML
收藏

通过增强式权重共享改进自动增强

Improving Auto-Augment via Augmentation-Wise Weight Sharing

SenseTime Research(商汤科技研究院);Beihang University(北京航空航天大学);University of Sydney(悉尼大学)

Keyu Tian, Chen Lin, Ming Sun, Luping Zhou, Junjie Yan, Wanli Ouyang

摘要:

自动搜索增强策略的最新进展大大提高了各种任务的性能。自动增强搜索的一个关键组成部分是对特定增强策略的评估过程,该过程用于返回奖励,通常运行数千次。简单的评估过程,包括完整的模型训练和验证,将是耗时的。为了提高效率,许多人选择牺牲评估的可靠性来换取速度。在本文中,我们深入研究了模型增强训练的动态。这启发我们设计一个基于增强权重共享(AWS)的强大而高效的代理任务,以优雅的方式形成快速而准确的评估过程。综合分析验证了该方法在有效性和效率方面的优越性。与现有的自动增强搜索方法相比,我们的方法发现的增强策略实现了更高的精度。在CIFAR-10上,我们实现了1.24%的top-1出错率,这是目前在没有额外训练数据的情况下表现最好的单一模型。在ImageNet上,我们得到ResNet-50的top-1出错率为20.36%,这使得基线增强的绝对出错率降低了3.34%。

英文摘要:

The recent progress on automatically searching augmentation policies has boosted the performance substantially for various tasks. A key component of automatic augmentation search is the evaluation process for a particular augmentation policy, which is utilized to return reward and usually runs thousands of times. A plain evaluation process, which includes full model training and validation, would be time-consuming. To achieve efficiency, many choose to sacrifice evaluation reliability for speed. In this paper, we dive into the dynamics of augmented training of the model. This inspires us to design a powerful and efficient proxy task based on the Augmentation-Wise Weight Sharing (AWS) to form a fast yet accurate evaluation process in an elegant way. Comprehensive analysis verifies the superiority of this approach in terms of effectiveness and efficiency. The augmentation policies found by our method achieve superior accuracies compared with existing auto-augmentation search methods. On CIFAR-10, we achieve a top-1 error rate of 1.24%, which is currently the best performing single model without extra training data. On ImageNet, we get a top-1 error rate of 20.36% for ResNet-50, which leads to 3.34% absolute error rate reduction over the baseline augmentation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑