We present a label noise self-filtering based learning method called ”NSFL” for improving generalizability of a classifier in label-noisy data. In this method, label noise is identified from normal samples by iteratively implementing the 2- means on loss values according to their different effects on loss values; and then, the label noise are filtered in a validation process. The NSFL does not rely on a specific loss function,resulting in a good performance in generalizability. Besides,it does not require to optimize any extra parameters of a specific measurement or noise estimation, so it is adaptive. In addition, it is proven that the learning process has the same convergence speed as the used loss function and is consistent with the optimal solution of the noise-free samples. To the best of authors knowledge, this is the first general and adaptive label noise-filtering method. The experimental results on synthetic and real datasets confirm that in comparison with the state-of-the-art methods, the proposed method is more effective in label noisy classification.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.