Research On Novel Partial Label Learning Algorithms

In recent years, learning with weak supervision has become one of the hottest research topics in machine learning community. Partial label learning is one of the important weakly-supervised learning frameworks. In partial label learning, an instance is associated with a set of candidate labels, and the true label of the instance is hidden in the candidate set. In partial label learning, the ground-truth label information is not available to the learning algorithm. To tackle with this problem, three approaches are proposed in this paper:An intuitive strategy to learn from partial label examples is to treat all candidate labels equally and make prediction by averaging their modeling outputs. Nonetheless, this strategy may suffer from the problem that the modeling output from the valid label is overwhelmed by those from the false positive labels. In this paper, an instance-based approach named IPAL is proposed by directly disambiguating the candidate label set, where an asymmetric weighted graph over the training examples is constructed by affinity relationship analysis. After that, I PAL tries to identify the valid label of each partial label example via an iterative label propagation procedure. During the testing phase, the unseen instance is classified based on minimum error reconstruction from its nearest neighbors.As one of the major machine learning techniques, maximum margin criterion has been employed to solve the partial label learning problem. Therein, disambiguation is performed by optimizing the margin between the maximum modeling output from candidate labels and that from non-candidate labels. However, in this formulation the margin between the ground-truth label and other candidate labels is not differentiated. In this paper, a new maximum margin formulation for partial label learning is proposed which aims to directly maximize the margin between the ground-truth label and all other labels.The common partial label learning strategy is to try to disambiguate the candidate labels, such as by identifying the ground-truth label iteratively or by treating each candidate label equally. Nevertheless, the above disambiguation strategy is prone to be misled by the false positive label(s) within candidate label set. In this paper, a new disambiguation-free approach to partial label learning is proposed by employing the well-known error-correcting output codes (ECOC) techniques. Specifically, to build the binary classifier with respect to each column coding, any partially labeled example will be regarded as a positive or negative training example only if its candidate label set entirely falls into the coding dichotomy. PL-ECOC transforms the partial label learning problem into a series of binary classification problems, and then predicts the label of unseen instance by combining the results of these binary classifiers.There are five chapters in this thesis. Definition, state-of-the-art and open research problems of partial label learning are introduced in Chapter 1. In Chapter 2 to Chapter 4, three partial label learning approaches named IPAL, M3P1 and PL-ECOC are introduced with their detailed experiment results. Finally, we conclude this paper in Chapter 5.