What is a Tabby? Interpretable Model Decisions by Learning Attribute-Based Classification Criteria

doi:10.1109/TPAMI.2019.2954501

	What is a Tabby? Interpretable Model Decisions by Learning Attribute-Based Classification Criteria
	Liu, Haomiao 1; Wang, Ruiping 2; Shan, Shiguang 2; Chen, Xilin 2
刊名	IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
	2021-05-01
卷号	43 期号:5 页码:1791-1807
关键词	Cats Prototypes Visualization Task analysis Streaming media Predictive models Scalability Interpretable model visual attributes convolutional neural network classification criteria
ISSN号	0162-8828
DOI	10.1109/TPAMI.2019.2954501
英文摘要	State-of-the-art classification models are usually considered as black boxes since their decision processes are implicit to humans. On the contrary, human experts classify objects according to a set of explicit hierarchical criteria. For example, "tabby is a domestic cat with stripes, dots, or lines", where tabby is defined by combining its superordinate category (domestic cat) and some certain attributes (e.g., has stripes). Inspired by this mechanism, we propose an interpretable Hierarchical Criteria Network (HCN) by additionally learning such criteria. To achieve this goal, images and semantic entities (e.g., taxonomies and attributes) are embedded into a common space, where each category can be represented by the linear combination of its superordinate category and a set of learned discriminative attributes. Specifically, a two-stream convolutional neural network (CNN) is elaborately devised, which embeds images and taxonomies with the two streams respectively. The model is trained by minimizing the prediction error of hierarchy labels on both streams. Extensive experiments on two widely studied datasets (CIFAR-100 and ILSVRC) demonstrate that HCN can learn meaningful attributes as well as reasonable and interpretable classification criteria. Therefore, the proposed method enables further human feedback for model correction as an additional benefit.
资助项目	973 Program[2015CB351802] ; Natural Science Foundation of China[61390511] ; Natural Science Foundation of China[61772500] ; Frontier Science Key Research Project CAS[QYZDJ-SSW-JSC009] ; Youth Innovation Promotion Association CAS[2015085]
WOS研究方向	Computer Science ; Engineering
语种	英语
出版者	IEEE COMPUTER SOC
WOS记录号	WOS:000637533800022
内容类型	期刊论文
源URL	[http://119.78.100.204/handle/2XEOYT63/16657]
专题	中国科学院计算技术研究所
通讯作者	Chen, Xilin
作者单位	1.Huawei EI Innovat Lab, Beijing 100085, Peoples R China 2.Chinese Acad Sci, Inst Comp Technol, CAS, Key Lab Intelligent Informat Proc, Beijing 100190, Peoples R China
推荐引用方式 GB/T 7714	Liu, Haomiao,Wang, Ruiping,Shan, Shiguang,et al. What is a Tabby? Interpretable Model Decisions by Learning Attribute-Based Classification Criteria[J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,2021,43(5):1791-1807.
APA	Liu, Haomiao,Wang, Ruiping,Shan, Shiguang,&Chen, Xilin.(2021).What is a Tabby? Interpretable Model Decisions by Learning Attribute-Based Classification Criteria.IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,43(5),1791-1807.
MLA	Liu, Haomiao,et al."What is a Tabby? Interpretable Model Decisions by Learning Attribute-Based Classification Criteria".IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 43.5(2021):1791-1807.