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 2403.13293cs.CVcs.AIcs.LG
收藏

使用可解释知识构建最优神经架构

Building Optimal Neural Architectures using Interpretable Knowledge

University of Alberta(阿尔伯塔大学);Huawei Technologies Canada(华为技术加拿大公司);Huawei Kirin Solution(华为麒麟解决方案部门)

Keith G. Mills, Fred X. Han, Mohammad Salameh, Shengyao Lu, Chunhua Zhou, Jiao He, Fengyu Sun, Di Niu

AI总结:

提出 AutoBuild 方案,通过对齐操作与架构模块的潜在嵌入与真实性能来分配可解释的重要性分数,无需搜索即可构建高性能神经架构,并能缩小搜索空间找到更优模型。

摘要:

神经架构搜索是一项成本高昂的实践。搜索空间可以涵盖大量的设计选择,而每次架构评估都需要不小的开销,这使得算法很难充分探索候选网络。在本文中,我们提出了 AutoBuild,这是一种学习将操作和架构模块的潜在嵌入与它们所出现的架构的真实性能对齐的方案。通过这样做,AutoBuild 能够为架构模块分配可解释的重要性分数,例如单个操作特征和更大的宏观操作序列,从而无需任何搜索即可构建高性能的神经网络。通过在最先进的图像分类、分割和 Stable Diffusion 模型上进行的实验,我们表明,通过挖掘相对较小的已评估架构集,AutoBuild 可以学会直接构建高质量架构,或帮助缩小搜索空间以聚焦于相关区域,从而找到优于原始标记架构和搜索基线所找到架构的更好架构。代码可在 https://github.com/Ascend-Research/AutoBuild 获取。

英文摘要:

Neural Architecture Search is a costly practice. The fact that a search space can span a vast number of design choices with each architecture evaluation taking nontrivial overhead makes it hard for an algorithm to sufficiently explore candidate networks. In this paper, we propose AutoBuild, a scheme which learns to align the latent embeddings of operations and architecture modules with the ground-truth performance of the architectures they appear in. By doing so, AutoBuild is capable of assigning interpretable importance scores to architecture modules, such as individual operation features and larger macro operation sequences such that high-performance neural networks can be constructed without any need for search. Through experiments performed on state-of-the-art image classification, segmentation, and Stable Diffusion models, we show that by mining a relatively small set of evaluated architectures, AutoBuild can learn to build high-quality architectures directly or help to reduce search space to focus on relevant areas, finding better architectures that outperform both the original labeled ones and ones found by search baselines. Code available at https://github.com/Ascend-Research/AutoBuild

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。

↑