Loading...

Probabilistic Approach for Multi-label Classification

Hosseini Akbarnejad, Amir Hossein | 2016

1876 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 49030 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Soleymani, Mahdie
  7. Abstract:
  8. In machine learning, classification is of great importance. Unlike the traditional single-label classification in which one instance can have only one label, in multi-label classification tasks, an instance can be associated with a set of labels. Multi-label classifiers have to address many problems including: considering correlations between labels, handling large-scale datasets with many instances and a large set of labels, and having only a fraction of valid label assignments in the training set. To tackle datasets with a large set of labels, recently embedding-based methods have been proposed which seek to represent the label assignments in an intermediate space. Subsequently, given the feature vectors, these methods seek to predict intermediate space representations, and then seek to transform intermediate space representations to label vectors by a proper mapping. Many state-of-the-art embedding-based methods use a linear mapping to represent the label vectors in an intermediate space. However, by doing so, these methods actually neglect the labels that are infrequently assigned to instances. In this research, we propose probabilistic methods which use a set of stochastic functions to model the relationship between the feature space, the intermediate space, and the label space. Moreover, these methods address the problem of missing labels, having many instances in the training set, and exploiting unlabeled instances. Furthermore, the proposed method of this research allows for exploiting the feature vectors of the tags. Experiments on real-world datasets show that the proposed methods of this research outperform state-of-the-art multi-label classifiers
  9. Keywords:
  10. Probabilistic Methods ; Multi-Label Classification ; Machine Learning ; Latent Label Space ; Missing Label

 Digital Object List

 Bookmark

...see more