Label Estimation Method with Modifications for Unreliable Examples in Taming

Methods for improving learning accuracy by utilizing a plurality of data sets having different reliabilities have been extensively studied. Unreliable data sets often include data with incorrect labels, and it is known that learning from such datasets adversely affects learning accuracy. In our study we focused on the learning problem, Taming which deals with two kinds of data sets with different reliabilities. We propose a label estimation method for use in data sets that include the data with incorrect labels. The proposed method is an extension of BaggTaming, which is proposed as a solution to Taming. We conducted experiments to verify the effectiveness of the proposed method using a benchmark data set which included the data with intentionally changed labels to incorrect. We confirmed that the learning accuracy can be improved by the proposed method when learning by using the data sets with modified labels.