人工智能培训

搜索

机器学习论文:硬度意识深度度量学习(Hardness-Aware Deep Metric Learning)

[复制链接]
rickyc 发表于 2019-3-15 12:58:41 | 显示全部楼层 |阅读模式
rickyc 2019-3-15 12:58:41 366 0 显示全部楼层
机器学习论文:硬度意识深度度量学习(Hardness-Aware Deep Metric Learning)本文提出了一种硬度感知深度量度学习(HDML)框架。大多数先前的深度量度学习方法采用硬负面挖掘策略来减轻缺乏信息样本的培训。然而,该挖掘策略仅利用训练数据的子集,这可能不足以全面地表征嵌入空间的整体几何形状。为了解决这个问题,我们进行线性插值嵌入以自适应地操纵它们的硬水平,并为循环训练生成相应的标签保留合成物,从而可以充分利用所有样本中隐藏的信息,并且该度量总是受到适当困难的挑战。我们的方法在广泛使用的CUB-200-2011,Cars196和Stanford OnlineProducts数据集上实现了极具竞争力的性能。
This paper presents a hardness-aware deep metric learning (HDML) framework.Most previous deep metric learning methods employ the hard negative miningstrategy to alleviate the lack of informative samples for training.However,this mining strategy only utilizes a subset of training data, which may not beenough to characterize the global geometry of the embedding spacecomprehensively.To address this problem, we perform linear interpolation onembeddings to adaptively manipulate their hard levels and generatecorresponding label-preserving synthetics for recycled training, so thatinformation buried in all samples can be fully exploited and the metric isalways challenged with proper difficulty.Our method achieves very competitiveperformance on the widely used CUB-200-2011, Cars196, and Stanford OnlineProducts datasets.机器学习论文:硬度意识深度度量学习(Hardness-Aware Deep Metric Learning) HFuUUFXU3Y8yqOpt.jpg
URL地址:https://arxiv.org/abs/1903.05503     ----pdf下载地址:https://arxiv.org/pdf/1903.05503    ----机器学习论文:硬度意识深度度量学习(Hardness-Aware Deep Metric Learning)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

rickyc当前离线
新手上路

查看:366 | 回复:0

快速回复 返回顶部 返回列表