TY - GEN
T1 - Relax DARTS
T2 - 18th Chinese Conference on Biometric Recognition, CCBR 2024
AU - Zhu, Hongyu
AU - Jin, Xin
AU - Liao, Hongchao
AU - Xiang, Yan
AU - El-Yacoubi, Mounim A.
AU - Qin, Huafeng
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.
PY - 2025/1/1
Y1 - 2025/1/1
N2 - Deep learning methods have shown good performance in the field of eye movement biometrics, but their network architecture relies on manual design and combined priori knowledge. To address these issues, we introduce automated network search (NAS) algorithms and present Relax DARTS, which is an improvement of the Differentiable Architecture Search (DARTS) to realize more efficient eye movement recognition network search and training. The key idea is to circumvent the issue of weight sharing by independently training the architecture parameters α to achieve a more precise target architecture. Moreover, the introduction of module input weights β allows cells the flexibility to select inputs, to alleviate the overfitting phenomenon and improve the model performance. Results on four public databases demonstrate that the Relax DARTS achieves state-of-the-art recognition performance. Notably, Relax DARTS exhibits adaptability to other multi-feature temporal classification tasks.
AB - Deep learning methods have shown good performance in the field of eye movement biometrics, but their network architecture relies on manual design and combined priori knowledge. To address these issues, we introduce automated network search (NAS) algorithms and present Relax DARTS, which is an improvement of the Differentiable Architecture Search (DARTS) to realize more efficient eye movement recognition network search and training. The key idea is to circumvent the issue of weight sharing by independently training the architecture parameters α to achieve a more precise target architecture. Moreover, the introduction of module input weights β allows cells the flexibility to select inputs, to alleviate the overfitting phenomenon and improve the model performance. Results on four public databases demonstrate that the Relax DARTS achieves state-of-the-art recognition performance. Notably, Relax DARTS exhibits adaptability to other multi-feature temporal classification tasks.
KW - Differentiable Architecture Search
KW - Eye movement biometrics
UR - https://www.scopus.com/pages/publications/85219214432
U2 - 10.1007/978-981-96-1071-6_11
DO - 10.1007/978-981-96-1071-6_11
M3 - Conference contribution
AN - SCOPUS:85219214432
SN - 9789819610709
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 112
EP - 122
BT - Biometric Recognition - 18th Chinese Conference, CCBR 2024, Proceedings
A2 - Yu, Shiqi
A2 - Jia, Wei
A2 - Shu, Xiangbo
A2 - Tang, Jinhui
A2 - Yuan, Xiaotong
A2 - Shan, Caifeng
A2 - Gui, Jie
A2 - Liu, Qingshan
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 22 November 2024 through 24 November 2024
ER -