摘要
多示例学习作为一种弱监督学习,被广泛应用于生物特征识别、图像检索及自然语言处理等领域。为解决现有多示例学习(multiple instance learning,MIL)的池化算子中缺乏可解释性和灵活性等问题,提出一种可学习的权重分配机制模型。通过示例得分得到示例权重,将示例得分与示例权重进行池化,求出包的表示,通过示例得分对权重分配函数进行学习。试验结果表明:文中方法具有更高的分类精度、较好的可解释性和灵活性。
Multiple instance learning as a weakly supervised learning method is widely used in the fields of the biometric recognition,image indexes and natural language processing.To solve the lack of interpretability and flexibility of the existing MIL pooling operators,an allocation mechanism model with learnable weights is proposed in this paper.The model has two parts which instance weights are obtained by instance scores and the bag representation is solved by pooling instance scores and instance weights.The weight distribution function is learned through instance scores.The experimental results show that the proposed method has higher classification accuracy,better interpretability and flexibility than other methods.
作者
吴文晴
袁立明
温显斌
徐海霞
WU Wenqing;YUAN Liming;WEN Xianbin;XU Haixia(School of Computer Science and Engineering,Tianjin University of Technology,Tianjin 300384,China;Key Laboratory of Computer Vision and System,Ministry of Education,Tianjin University of Technology,Tianjin 300384,China;Tianjin Key Laboratory of Intelligence Computing and Novel Software Technology,Tianjin University of Technology,Tianjin 300384,China)
出处
《天津理工大学学报》
2023年第5期29-35,共7页
Journal of Tianjin University of Technology
基金
国家自然科学基金(61472278)
天津市新一代人工智能重大专项(18ZXZNGX00150)
天津市自然科学基金(18JCYBJC84800)
天津市教委自然科学重点项目(2017ZD13)
天津市教委自然科学一般项目(2017KJ255)
天津市技术创新引导专项项目(21YDTPJC00250)
天津市新一代人工智能科技重大专项项目(18ZXZNGX00150)。
关键词
模式识别
深度学习
多示例学习
权重
pattern recognition
deep learning
multiple instance learning
weight