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 2009.09209cs.CVcs.AI
收藏

MSR-DARTS:可微体系结构搜索的最小稳定秩

MSR-DARTS: Minimum Stable Rank of Differentiable Architecture Search

School of Computing, Tokyo Institute of Technology(东京工业大学计算学院);Graduate School of Information Science and Technology, The University of Tokyo(东京大学情报理工学系研究科);Center for Advanced Intelligence Project, RIKEN(理化学研究所先进智能项目中心)

Kengo Machida, Kuniaki Uto, Koichi Shinoda, Taiji Suzuki

摘要:

在神经结构搜索(NAS)中,可微结构搜索(DARTS)由于其高效性而受到广泛关注。它定义了一个具有混合边的过参数化网络,每个边代表所有候选算子,并以交替的方式联合优化网络及其架构的权重。然而,这种方法发现一个模型的权重收敛速度比其他的,这样一个模型的收敛速度最快,往往导致过拟合。因此,由此产生的模型不能总是很好地推广。为了克服这个问题,我们提出了一种方法称为最小稳定秩DARTS(MSR-DARTS),找到一个最好的泛化误差的模型,取代结构优化与选择过程中使用的最小稳定秩标准。具体地,卷积算子由矩阵表示,并且MSR-DARTS选择具有最小稳定秩的卷积算子。我们在CIFAR-10和ImageNet数据集上评估了MSR-DARTS。它在CIFAR-10上的0.3 GPU天内实现了2.54%的出错率和4.0M参数,在ImageNet上的前1出错率为23.9%。官方代码可在https://github.com/mtaecchhi/msrdarts.git上获得。

英文摘要:

In neural architecture search (NAS), differentiable architecture search (DARTS) has recently attracted much attention due to its high efficiency. It defines an over-parameterized network with mixed edges, each of which represents all operator candidates, and jointly optimizes the weights of the network and its architecture in an alternating manner. However, this method finds a model with the weights converging faster than the others, and such a model with fastest convergence often leads to overfitting. Accordingly, the resulting model cannot always be well-generalized. To overcome this problem, we propose a method called minimum stable rank DARTS (MSR-DARTS), for finding a model with the best generalization error by replacing architecture optimization with the selection process using the minimum stable rank criterion. Specifically, a convolution operator is represented by a matrix, and MSR-DARTS selects the one with the smallest stable rank. We evaluated MSR-DARTS on CIFAR-10 and ImageNet datasets. It achieves an error rate of 2.54% with 4.0M parameters within 0.3 GPU-days on CIFAR-10, and a top-1 error rate of 23.9% on ImageNet. The official code is available at https://github.com/mtaecchhi/msrdarts.git.

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。

↑