Logit Mixing Training for More Reliable and Accurate Prediction
- Authors
- Bang, D.[Bang, D.]; Baek, K.[Baek, K.]; Kim, J.[Kim, J.]; Jeon, Y.[Jeon, Y.]; Kim, J.-H.[Kim, J.-H.]; Kim, J.[Kim, J.]; Lee, J.[Lee, J.]; Shim, H.[Shim, H.]
- Issue Date
- 2022
- Publisher
- International Joint Conferences on Artificial Intelligence
- Citation
- IJCAI International Joint Conference on Artificial Intelligence, pp.2812 - 2819
- Indexed
- SCOPUS
- Journal Title
- IJCAI International Joint Conference on Artificial Intelligence
- Start Page
- 2812
- End Page
- 2819
- URI
- https://scholarworks.bwise.kr/skku/handle/2021.sw.skku/100903
- ISSN
- 1045-0823
- Abstract
- When a person solves the multi-choice problem, she considers not only what is the answer but also what is not the answer. Knowing what choice is not the answer and utilizing the relationships between choices, she can improve the prediction accuracy. Inspired by this human reasoning process, we propose a new training strategy to fully utilize inter-class relationships, namely LogitMix. Our strategy is combined with recent data augmentation techniques, e.g., Mixup, Manifold Mixup, CutMix, and PuzzleMix. Then, we suggest using a mixed logit, i.e., a mixture of two logits, as an auxiliary training objective. Since the logit can preserve both positive and negative inter-class relationships, it can impose a network to learn the probability of wrong answers correctly. Our extensive experimental results on the image- and language-based tasks demonstrate that LogitMix achieves state-of-the-art performance among recent data augmentation techniques regarding calibration error and prediction accuracy. © 2022 International Joint Conferences on Artificial Intelligence. All rights reserved.
- Files in This Item
- There are no files associated with this item.
- Appears in
Collections - Computing and Informatics > Computer Science and Engineering > 1. Journal Articles
![qrcode](https://api.qrserver.com/v1/create-qr-code/?size=55x55&data=https://scholarworks.bwise.kr/skku/handle/2021.sw.skku/100903)
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.