Candidate Label-aware Similarity Graph for Partial Label Data

In partial label learning, the label of instance is represented by a label set, among which only one is the true label. The existing partial label learning algorithm only use features of instances to measure the similarity, and lacks of utilization of information hidden in the label set. In this paper, we proposes a candidate label-aware similarity graph constructing method for partial label data which effectively combines candidate label information using Jaccard distance and linear reconstruction to measure the similarity between instances. First, we calculate the similarity between instances and candidate label-sets. Then, the similarity graph is constructed by combining the instance similarity and the label-set similarity. Extensive experiments on artificial as well as real-world PL data sets show that our method can measure the similarity of partial-label data better and improve the performance of graph-based partial label learning algorithm obviously.

[1]  Francesco Orabona,et al.  Learning from Candidate Labeling Sets , 2010, NIPS.

[2]  Naonori Ueda,et al.  Mobile Activity Recognition through Training Labels with Inaccurate Activity Segments , 2016, MobiQuitous.

[3]  Bo An,et al.  Leveraging Latent Label Distributions for Partial Label Learning , 2018, IJCAI.

[4]  Min-Ling Zhang,et al.  Confidence-Rated Discriminative Partial Label Learning , 2017, AAAI.

[5]  Ben Taskar,et al.  Learning from Partial Labels , 2011, J. Mach. Learn. Res..

[6]  Zhi-Hua Zhou,et al.  A brief introduction to weakly supervised learning , 2018 .

[7]  Xu-Ying Liu,et al.  Partial Label Learning via Feature-Aware Disambiguation , 2016, KDD.

[8]  Fei Yu,et al.  Solving the Partial Label Learning Problem: An Instance-Based Approach , 2015, IJCAI.

[9]  Fei Yu,et al.  Disambiguation-Free Partial Label Learning , 2017, IEEE Trans. Knowl. Data Eng..

[10]  Fei Yu,et al.  Maximum margin partial label learning , 2017, Machine Learning.

[11]  Dong Xu,et al.  Learning by Associating Ambiguously Labeled Images , 2013, 2013 IEEE Conference on Computer Vision and Pattern Recognition.

[12]  Eyke Hüllermeier,et al.  Learning from ambiguously labeled examples , 2005, Intell. Data Anal..

[13]  Xiaoli Z. Fern,et al.  Rank-loss support instance machines for MIML instance annotation , 2012, KDD.

[14]  Thomas G. Dietterich,et al.  A Conditional Multinomial Mixture Model for Superset Label Learning , 2012, NIPS.

[15]  Cordelia Schmid,et al.  Multiple Instance Metric Learning from Automatically Labeled Bags of Faces , 2010, ECCV.