Showing 1–1 of 1 results for author: von Rekowski, C S
-
Probabilistic Active Learning for Active Class Selection
Authors:
Daniel Kottke,
Georg Krempl,
Marianne Stecklina,
Cornelius Styp von Rekowski,
Tim Sabsch,
Tuan Pham Minh,
Matthias Deliano,
Myra Spiliopoulou,
Bernhard Sick
Abstract:
In machine learning, active class selection (ACS) algorithms aim to actively select a class and ask the oracle to provide an instance for that class to optimize a classifier's performance while minimizing the number of requests. In this paper, we propose a new algorithm (PAL-ACS) that transforms the ACS problem into an active learning task by introducing pseudo instances. These are used to estimat…
▽ More
In machine learning, active class selection (ACS) algorithms aim to actively select a class and ask the oracle to provide an instance for that class to optimize a classifier's performance while minimizing the number of requests. In this paper, we propose a new algorithm (PAL-ACS) that transforms the ACS problem into an active learning task by introducing pseudo instances. These are used to estimate the usefulness of an upcoming instance for each class using the performance gain model from probabilistic active learning. Our experimental evaluation (on synthetic and real data) shows the advantages of our algorithm compared to state-of-the-art algorithms. It effectively prefers the sampling of difficult classes and thereby improves the classification performance.
△ Less
Submitted 9 August, 2021;
originally announced August 2021.
