Cascade-Structured Classifier Based on Adaptive Devices

This paper presents a novel approach to decision making based on uncertain data. Typical supervised learning algorithms assume that training data is perfectly accurate, and weight each training instance equally, resulting in a static classifier, whose structure can not be changed once built unless retrained from scratch. In this paper, we address this issue by using adaptive devices that can be incrementally trained, allowing them to aggregate new pieces of information while processing new input entries. We also propose a confidence model to weight each instance according to an estimate of its likelihood.