Fraunhofer-Gesellschaft

Publica

Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Comparing visual-interactive labeling with active learning: An experimental study

 
: Bernard, Jürgen; Hutter, Marco; Zeppelzauer, Matthias; Fellner, Dieter W.; Sedlmair, Michael

:

IEEE transactions on visualization and computer graphics 24 (2018), Nr.1, S.298-308
ISSN: 1077-2626
Englisch
Zeitschriftenaufsatz
Fraunhofer IGD ()
Labeling; information visualization; visual analytic; Active Learning; machine learning; classification; Guiding Theme: Digitized Work; Guiding Theme: Individual Health; Research Area: Human computer interaction (HCI); Research Area: Modeling (MOD)

Abstract
Labeling data instances is an important task in machine learning and visual analytics. Both fields provide a broad set of labeling strategies, whereby machine learning (and in particular active learning) follows a rather model-centered approach and visual analytics employs rather user-centered approaches (visual-interactive labeling). Both approaches have individual strengths and weaknesses.
In this work, we conduct an experiment with three parts to assess and compare the performance of these different labeling strategies. In our study, we (1) identify different visual labeling strategies for user-centered labeling, (2) investigate strengths and weaknesses of labeling strategies for different labeling tasks and task complexities, and (3) shed light on the effect of using different visual encodings to guide the visual-interactive labeling process. We further compare labeling of single versus multiple instances at a time, and quantify the impact on efficiency. We systematically compare the performance of visual interactive labeling with that of active learning. Our main findings are that visual-interactive labeling can outperform active learning, given the condition that dimension reduction separates well the class distributions. Moreover, using dimension reduction in combination with additional visual encodings that expose the internal state of the learning model turns out to improve the performance of visual-interactive labeling.

: http://publica.fraunhofer.de/dokumente/N-477628.html