Ensemble-Based Out-of-Distribution Detection

Autor: Iksoo Shin, Kien Mai Ngoc, Kyong-Ha Lee, Donghun Yang, Myunggwon Hwang
Jazyk: angličtina
Rok vydání: 2021
Předmět:
Computer Networks and Communications
Computer science
Gaussian
Feature vector
lcsh:TK7800-8360
ComputerApplications_COMPUTERSINOTHERSYSTEMS
02 engineering and technology
030218 nuclear medicine & medical imaging
Image (mathematics)
03 medical and health sciences
symbols.namesake
0302 clinical medicine
Classifier (linguistics)
0202 electrical engineering
electronic engineering
information engineering
Electrical and Electronic Engineering
Mahalanobis distance
business.industry
Deep learning
distance metric learning
lcsh:Electronics
deep learning
Pattern recognition
Siamese network
Class (biology)
triplet network
ensemble method
Hardware and Architecture
Control and Systems Engineering
Signal Processing
Metric (mathematics)
Softmax function
symbols
020201 artificial intelligence & image processing
Artificial intelligence
confidence score
business
out-of-distribution detection
Zdroj: Electronics, Vol 10, Iss 567, p 567 (2021)
Electronics
Volume 10
Issue 5
ISSN: 2079-9292
Popis: To design an efficient deep learning model that can be used in the real-world, it is important to detect out-of-distribution (OOD) data well. Various studies have been conducted to solve the OOD problem. The current state-of-the-art approach uses a confidence score based on the Mahalanobis distance in a feature space. Although it outperformed the previous approaches, the results were sensitive to the quality of the trained model and the dataset complexity. Herein, we propose a novel OOD detection method that can train more efficient feature space for OOD detection. The proposed method uses an ensemble of the features trained using the softmax-based classifier and the network based on distance metric learning (DML). Through the complementary interaction of these two networks, the trained feature space has a more clumped distribution and can fit well on the Gaussian distribution by class. Therefore, OOD data can be efficiently detected by setting a threshold in the trained feature space. To evaluate the proposed method, we applied our method to various combinations of image datasets. The results show that the overall performance of the proposed approach is superior to those of other methods, including the state-of-the-art approach, on any combination of datasets.
Databáze: OpenAIRE
Externí odkaz: