Research On Multiple Instance Leanring Based On Extreme Learning Machine

In many fields, it is found that many of the problems are essentially formulated making use of the multiple instance learning(MIL) setting. A few fields that use this framework can be mentioned, such as drug discovery, image categorization, object detection, text categorization, speaker identification. Thus, MIL problem becomes an important topic in the machine leaning field. Numerous multiple instance algorithms have been published over the last decade.Since real-world data sets usually contain large instances, most existed learning algorithms suffer from slow training speed. It is meaningful to develop efficient and effective MIL algorithm. In this thesis, we propose two MIL methods based on the extreme learning machine(ELM). The main contribution of this thesis is as follows:1. As a learning paradigm, MIL is different from traditional supervised learning that handles the classification of bags comprising unlabeled instances. In this thesis, a novel efficient method based on extreme learning machine(ELM) is proposed to address MIL problem. First, the most qualified instance is selected in each bag through a single hidden layer feed forward network(SLFN) whose input and output weights are both initialed randomly, and the single selected instance is used to represent every bag. Second, the modified ELM model is trained by using the selected instances to update the output weights. Experiments on several benchmark data sets and multiple instance regression data sets show that the ELM-MIL achieves good performance; moreover, it runs several times or even hundreds of times faster than other similar MIL algorithms.2. In order to overcome MIL-ELM’s instability, bagging is exploited to enhance MIL-ELM stability. Experiments show that the stability as well as the test accuracy of MIL-ELM are increased with bagging.3. This thesis presents a novel MIL algorithm for an extreme learning machine called MI-ELM. A radial basis kernel extreme learning machine is adapted to approach the MIL problem using Hausdorff distance to measure the distance between the bags. The clusters in the hidden layer are composed of bags that are randomly generated. Because we do not need to tune the parameters for the hidden layer, MI-ELM can learn very fast. The experimental results on classifications and multiple-instance regression data sets demonstrate that the MI-ELM is useful and efficient as compared to the state-of-the-art algorithms.4. The proposed algorithms were successfully applied in drug activity prediction, and achieve good performance. Based on the improved Kmeans clustering algorithm for image segmentation,the MI-ELM and ELM-MIL are evaluated on image categorization. The results show that the proposed MIL algorithms achieve competitive performance among the state of the art. Finally, the methods are tested on multiple instance regression data sets, and shown to provide results comparable to previous best MIL methods.