Unsupervised Outlier Detection and Semi-Supervised Learning ; CU-CS-976-04
Adam Vinueza, Gregory Z. Grudić · CU Scholar (University of Colorado Boulder) · 2004
A familiar problem in machine learning is to determine which data points are outliers when the underlying distribution is unknown.In this paper, we adapt a simple algorithm from Zhou et al [3], designed for semisupervised learning, and show that it not only can automatically detect outliers by using local and global consistency of data points, but also automatically select optimal learning parameters, as well as predict class outliers for points introduced after training.