Kernel K-means Based Framework for Aggregate Outputs Classification

Abstract

Aggregate outputs learning is a newly proposed setting in data mining and machine learning. It differs from the classical supervised learning setting in that, training samples are packed into bags with only the aggregate outputs (labels for classification or real values for regression) provided. This problem is associated with several kinds of application background. We focus on the aggregate outputs classification problem in this paper, and set up a framework based on kernel K-means to solve it. Two concrete algorithms based on our framework are proposed, each of which can cope with both binary and multi-class scenarios. The experimental results suggest that our algorithms outperform the state-of-art technique. Also, we propose a new setting for patch extraction in the Content Based Image Retrieval procedure by using the algorithm.