Abstract
In multi-label learning, each object is represented by a single instance and associated with multiple labels simultaneously. Existing multi-label learning approaches mainly construct classification models with a fixed set of target labels ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">observed labels</i> ). However, in the big data era, it is difficult to provide a fully complete label set for a data set. In some real applications, there are multiple labels <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">hidden</i> in the data set, especially for those large-scale data sets. In this paper, a novel approach named MLLHL is proposed to not only discover the hidden labels in the training data but also predict these <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">hidden labels</i> and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">observed labels</i> for unseen examples simultaneously. We assume that the observed labels are just a subset of labels which are selected from the full label set, and the rest ones are omitted by the annotators during the labeling stage. Extensive experiments show the competitive performance of MLLHL against other state-of-the-art multi-label learning approaches.
Highlights
As a popular machine learning task, multi-label learning [1], [2] aims to learn a robust model based on the training data which can predict a set of possible labels for new data instances
We assume that the observed label matrix Y is a subset of l different columns selected from F, and the rest q columns are omitted by the annotators
We can find that the highest computational costs are Eigenvalue decomposition in Step 4, the inverse and singular value decomposition (SVD) operations in Steps 5 and 7
Summary
As a popular machine learning task, multi-label learning [1], [2] aims to learn a robust model based on the training data which can predict a set of possible labels for new data instances. While existing multi-label learning approaches mainly construct classification models for the fixed set of observed labels. In the data, they will be fully neglected, especially for a large-scale data set with huge number of class labels These hidden labels may provide novel and useful knowledge for the learning task. A novel approach named MLLHL (i.e., MultiLabel Learning with Hidden Labels) is proposed. It can discover the hidden labels in the training data and predict them as well as the other observed labels for new examples simultaneously.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
