ABSTRACT

Uncertainty is a common phenomenon in machine learning, which can be found in every phase of learning, such as data preprocessing, algorithm design, and model selection. The representation, measurement, and handling of uncertainty have a significant impact on the performance of a learning system. There are four common uncertainties in machine learning, that is, randomness [1], fuzziness [2], roughness [3], and nonspecificity [4]. In this chapter, we mainly introduce the first three kinds of uncertainty, briefly list the fourth uncertainty, and give a short discussion about the relationships among the four uncertainties.