Learning Outlier-Aware Representation with Synthetic Boundary Samples

Jen Tzung Chien*, Kuan Chen

*此作品的通信作者

研究成果: Conference contribution同行評審

摘要

Out-of-distribution (OOD) detection provides an essential scheme to build a reliable deep learning system. Some of the previous works additionally acquired OOD samples for model training or hyperparameter tuning. However, OOD samples are usually missing in real-world scenarios and the prior knowledge of these samples is commonly unavailable. To cope with these issues, this paper presents a novel approach to synthesizing the informative samples near the boundary between in-distribution and OOD which sufficiently reflect OOD samples in training stage. These synthetic boundary samples are located right outside in-distribution (ID) where the latent variable is distributed by a multivariate Gaussian. Accordingly, this paper presents the outlier-aware representation learning which utilizes the synthetic boundary samples to train an OOD detector by only using the unlabeled ID data. The model is then trained to learn a compact decision boundary between ID and OOD samples. The experiments demonstrate that the proposed methods outperform state-of-the-art performance in presence of different OOD data.

原文English
主出版物標題Proceedings of the 2023 IEEE 33rd International Workshop on Machine Learning for Signal Processing, MLSP 2023
編輯Danilo Comminiello, Michele Scarpiniti
發行者IEEE Computer Society
ISBN(電子)9798350324112
DOIs
出版狀態Published - 2023
事件33rd IEEE International Workshop on Machine Learning for Signal Processing, MLSP 2023 - Rome, Italy
持續時間: 17 9月 202320 9月 2023

出版系列

名字IEEE International Workshop on Machine Learning for Signal Processing, MLSP
2023-September
ISSN(列印)2161-0363
ISSN(電子)2161-0371

Conference

Conference33rd IEEE International Workshop on Machine Learning for Signal Processing, MLSP 2023
國家/地區Italy
城市Rome
期間17/09/2320/09/23

指紋

深入研究「Learning Outlier-Aware Representation with Synthetic Boundary Samples」主題。共同形成了獨特的指紋。

引用此