Home // ICAS 2020, The Sixteenth International Conference on Autonomic and Autonomous Systems // View article

In the Depths of Hyponymy: A Step Towards Lifelong Learning

Authors:
Tommaso Boccato
Timothy Patten
Markus Vincze
Stefano Ghidoni

Keywords: Classification; Lifelong learning; Open set learning

Abstract:
This paper proposes a novel framework for lifelong learning of semantic classes in order to extend the operational time of robots deployed in real-world and uncontrolled environments. In contrast to the common approach that assumes fixed object classes, the proposed framework keeps track of the intra-class variability over time in order to refine the class definition encoded into a classifier. A carefully designed metric is also presented to quantify the intra-class variability, which leads to automatic triggering of the class restructuring. Experiments performed with the CIFAR-100 dataset validate the framework and the measure of intra-class variability.

Pages: 103 to 109

Copyright: Copyright (c) IARIA, 2020

Publication date: September 27, 2020

Published in: conference

ISSN: 2308-3913

ISBN: 978-1-61208-787-0

Location: Lisbon, Portugal

Dates: from September 27, 2020 to October 1, 2020