Publication Details
Overview
 
 
Isel Grau, Dipankar Sengupta, Maria M. Garcia Lorenzo, Ann Nowe
 

Chapter in Book/ Report/ Conference proceeding

Abstract 

Semi-supervised classifiers combine labeled and unlabeleddata during the learning phase in order to increaseclassifier{\textquoteright}s generalization capability. However, most successfulsemi-supervised classifiers involve complex ensemble structuresand iterative algorithms which make it difficult to explain theoutcome, thus behaving like black boxes. Furthermore, duringan iterative self-labeling process, mistakes can be propagated ifno amending procedure is used. In this paper, we build upon aninterpretable self-labeling grey-box classifier that uses a blackbox to estimate the missing class labels and a white box to makethe final predictions. We propose a Rough Set based approach foramending the self-labeling process. We compare its performanceto the vanilla version of our self-labeling grey-box and theuse of a confidence-based amending. In addition, we introducesome measures to quantify the interpretability of our model.The experimental results suggest that the proposed amendingimproves accuracy and interpretability of the self-labeling grey-box,thus leading to superior results when compared to state-of-the-art semi-supervised classifiers.

Reference 
 
 
DOI