한국산업정보학회논문지 (Journal of Korea Society of Industrial Information Systems)

한국산업정보학회 (Korea Society of Industrial Information Systems)
권호 표지
DOI QR코드

DOI QR Code

입출력구조와 신경망 모델에 따른 딥러닝 기반 정규화 기법의 성능 분석

Performance Analysis of Deep Learning-based Normalization According to Input-output Structure and Neural Network Model

  • 류창수 (영남이공대학교) ;
  • 김근환 (세종대학교 해양시스템융합공학과)
  • Changsoo Ryu ;
  • Geunhwan Kim
  • 투고 : 2024.04.16
  • 심사 : 2024.07.29
  • 발행 : 2024.08.30
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

본 논문에서는 다양한 신경망 모델과 입출력 구조에 따른 정규화 기법의 성능을 비교 분석하였다. 분석을 위해 균등한 잡음과 최대 3개의 간섭 신호가 있는 잡음 환경에 대한 시뮬레이션 기반의 데이터 세트를 사용하였다. 실험 결과, 잡음 분산을 직접 출력하는 End-to-End 구조에 대해서 1-D 콘볼루션 신경망과 BiLSTM 모델을 사용할 경우 우수한 성능을 보였으며, 특히 간섭 신호에 대해 강건한 것으로 분석되었다. 이러한 결과는 다층 퍼셉트론 신경망과 트랜스포머보다 1-D 콘볼루션 신경망 및 BiLSTM 모델이 귀납적 편향이 강하기 때문에 나타난 것으로 판단된다. 이 논문의 분석 결과는 향후 딥러닝 기반 정규화 기법 연구에 유용한 기준점으로 활용될 수 있을 것으로 기대된다.

In this paper, we analyzed the performance of normalization according to various neural network models and input-output structures. For the analysis, a simulation-based dataset for noise environments with homogeneous and up to three interfering signals was used. As a result, the end-to-end structure that directly outputs noise variance showed superior performance when using a 1-D convolutional neural network and BiLSTM model, and was analyzed to be particularly robust against interference signals. This is because the 1-D convolutional neural network and bidirectional long short-term memory models have stronger inductive bias than the multilayer perceptron and transformer models. The analysis of this paper are expected to be used as a useful reference for future research on deep learning-based normalization.

키워드

과제정보

이 논문은 2023학년도 영남이공대학교 연구조성비 지원에 의한 것임.

참고문헌

  1. Abraham, D. A. (2019). Underwater Acoustic Signal Processing: Modeling, Detection, and Estimation, Springer. 
  2. Chen, Z., Zhao, X., Zhou, Z., Ma, X., Cheng, Q., Cai, X., Jiang, B., Khan, R., Sharma, P. K., Alfarraj, O. and Tolba, A. (2023). The Adaptive Constant False Alarm Rate for Sonar Target Detection based on Back Propagation Neural Network Access, IET Signal Processing, 17(3), 12196. 
  3. Feintuch, S., Permuter, H. H., Bilik, I. and Tabrikian, J. (2023). Neural Network-based Multi-target Detection within Correlated Heavy-tailed Clutter, IEEE Transactions on Aerospace and Electronic Systems, 59(5), 5684-5698. 
  4. Goodfellow, I., Bengio, Y. and Courville, A. (2016). Deep Learning. MIT Press.
  5. Hong, S.-W. and Han, D.-S. (2011). OSR CFAR Robust to Multiple Underwater Target Environments, Journal of the Institute of Electronics Engineers of Korea TC, 48(4), 47-52. 
  6. Kim, G. and Lee, D. (2022). Mutual Interference Suppression of the Sinusoidal Frequency Modulated Pulse using SHAPE Algorithm, Journal of Korea Society of Industrial Information Systems, 27(5), 49-59. 
  7. Kim, G., Lee, S., Lee, K. and Lee, D. (2020). A Study of Active Pulse Classification Algorithm using Multi-label Convolutional Neural Networks, Journal of Korea Society of Industrial Information Systems, 25(4), 29-38. 
  8. Kim, G., Yoon, K., Lee, D., Cho, C., Hong, J. and Lee, K. (2019). Design of the Robust Generalized Sinusoidal Frequency Modulated Pulse in Reverberation Environments, Journal of Korea Society of Industrial Information Systems, 24(5), 95-104. 
  9. Li, P., Wang, P., Berntorp, K. and Liu, H. (2022). Exploiting Temporal Relations on Radar Perception for Autonomous Driving, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 17071-17080. 
  10. Lin, C.-H., Lin, Y.-C., Bai, Y., Chung, W.-H., Lee, T.-S. and Huttunen, H. (2019). DL-CFAR: A Novel CFAR Target Detection Method based on Deep Learning, IEEE 90th Vehicular Technology Conference (VTC2019 - Fall), Sep. 22-25, Honolulu, Hawaii, USA, pp. 1-6. 
  11. Smith, M. E. and Varshney, P. K. (1997). VI-CFAR: A Novel CFAR Algorithm based on Data Variability, Proceedings of the 1997 IEEE National Radar Conference, May. 13-15, Syracuse, New York, USA, pp. 263-268. 
  12. Zeng, T., Zhang, T., Shao, Z., Xu, X., Zhang, W., Shi, J., Wei, S. and Zhang, X. (2024). CFAR-DP-FW: A CFAR-guided Dual-Polarization Fusion Framework for Large Scene SAR Ship Detection, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 7242-7259. 
  13. Zhao, J., Jiang, R., Wang, X. and Gao, H. (2019). Robust CFAR Detection for Multiple Targets in K-distributed Sea Clutter based on Machine Learning, Symmetry. 11(12), 1482. 
  14. Zhihui, C., Fang, W., Song, Y., He, L., Song, C. and Xu, Z. (2021). DNN-based Peak Sequence Classification CFAR Detection Algorithm for High-resolution FMCW Radar, IEEE Transactions on Geoscience and Remote Sensing, 60, 1-15.