TY - GEN
T1 - Transductive Feature Space Regularization for Few-shot Bioacoustic Event Detection
AU - Tan, Yizhou
AU - Ai, Haojun
AU - Li, Shengchen
AU - Zhang, Feng
N1 - Funding Information:
The research project is supported partly by the National Natural Science Foundation of China (No: 62001038 and No: 61971316), and Gusu Innovation and Entrepreneurship Leading Talents Programme - Youth Innovation Leading Talent (ZXL2022472). *Corresponding Author
Publisher Copyright:
© 2023 International Speech Communication Association. All rights reserved.
PY - 2023/8/20
Y1 - 2023/8/20
N2 - In few-shot bioacoustic event detection, besides interested target events, background noises and various uninterested sound events lead to complex decision boundaries, which require regularized feature distributions in feature space. Due to the low label availability of uncertain noise events, existing few-shot learning methods with entropy-based regularizers suffer from overfitting during optimization. In this paper, we propose a transductive inference model with a prior knowledge based regularizer (PKR) to overcome the overfitting problem. We use a task-adaptive feature extractor to reconstruct a regularized feature space. A PKR is proposed to minimize the divergence between the original and reconstructed feature space. The development set of DCASE 2022 Task 5 is adopted as the experimental dataset. With the increasing iterations, the proposed model performs with long-lasting results around 55.43 F-measure, and well solves the overfitting problem in transductive inference.
AB - In few-shot bioacoustic event detection, besides interested target events, background noises and various uninterested sound events lead to complex decision boundaries, which require regularized feature distributions in feature space. Due to the low label availability of uncertain noise events, existing few-shot learning methods with entropy-based regularizers suffer from overfitting during optimization. In this paper, we propose a transductive inference model with a prior knowledge based regularizer (PKR) to overcome the overfitting problem. We use a task-adaptive feature extractor to reconstruct a regularized feature space. A PKR is proposed to minimize the divergence between the original and reconstructed feature space. The development set of DCASE 2022 Task 5 is adopted as the experimental dataset. With the increasing iterations, the proposed model performs with long-lasting results around 55.43 F-measure, and well solves the overfitting problem in transductive inference.
KW - Bioacoustic Event Detection
KW - Few-shot Learning
KW - Transductive Inference
UR - http://www.scopus.com/inward/record.url?scp=85171551514&partnerID=8YFLogxK
U2 - 10.21437/Interspeech.2023-579
DO - 10.21437/Interspeech.2023-579
M3 - Conference Proceeding
AN - SCOPUS:85171551514
VL - 2023-August
T3 - Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH
SP - 571
EP - 575
BT - Proceeding of INTERSPEECH 2023
T2 - 24th International Speech Communication Association, Interspeech 2023
Y2 - 20 August 2023 through 24 August 2023
ER -