TY - GEN
T1 - Event-Triggered Pseudo Supervised Meta Learning for Susceptibility Assessment of Acute Mountain Sickness
AU - Wang, Lei
AU - Shi, Dawei
AU - Zhu, Lingling
AU - Wang, Junzheng
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - Acute altitude sickness presents as a syndrome characterized by individuals' incapacity to adapt to the hypoxic conditions of high-altitude environments. The assessment of this sickness typically relies on the subjective evaluation of symptom severity using the Lake Louise score, introducing potential subjective biases and rendering the labels unreliable. To address this challenge, this paper introduces a weakly supervised meta-learning algorithm tailored for the assessment of susceptibility to acute altitude sickness within a few-shot framework. Initially, objective sleep data is subjected to clustering, and pseudo-labels are generated from the cluster result exhibiting the highest mutual information with the Lake Louise score. This approach capitalizes on the mutual complementarity of subjective and objective data, thereby enhancing label credibility. Subsequently, a task-driven pseudo-label supervised meta-learning algorithm is formulated to comprehensively explore individual differences. Moreover, recognizing the necessity for model updates during online learning, an event-triggered online learning algorithm is devised to optimize computational resources. Experimental validation utilizing data from 18 real subjects demonstrates promising results: an accuracy of 79.900% and an F1 score of 79.342% for acute altitude sickness susceptibility assessment, with further improvements evident through the online updated model's accuracy of 81.992% and F1 score of 81.366% on the test set. These outcomes underscore the viability of the proposed event-triggered weakly supervised meta-learning algorithm in the assessment of susceptibility to acute altitude sickness.
AB - Acute altitude sickness presents as a syndrome characterized by individuals' incapacity to adapt to the hypoxic conditions of high-altitude environments. The assessment of this sickness typically relies on the subjective evaluation of symptom severity using the Lake Louise score, introducing potential subjective biases and rendering the labels unreliable. To address this challenge, this paper introduces a weakly supervised meta-learning algorithm tailored for the assessment of susceptibility to acute altitude sickness within a few-shot framework. Initially, objective sleep data is subjected to clustering, and pseudo-labels are generated from the cluster result exhibiting the highest mutual information with the Lake Louise score. This approach capitalizes on the mutual complementarity of subjective and objective data, thereby enhancing label credibility. Subsequently, a task-driven pseudo-label supervised meta-learning algorithm is formulated to comprehensively explore individual differences. Moreover, recognizing the necessity for model updates during online learning, an event-triggered online learning algorithm is devised to optimize computational resources. Experimental validation utilizing data from 18 real subjects demonstrates promising results: an accuracy of 79.900% and an F1 score of 79.342% for acute altitude sickness susceptibility assessment, with further improvements evident through the online updated model's accuracy of 81.992% and F1 score of 81.366% on the test set. These outcomes underscore the viability of the proposed event-triggered weakly supervised meta-learning algorithm in the assessment of susceptibility to acute altitude sickness.
KW - Acute Mountain Sickness
KW - Event-triggered
KW - Pseudo Supervised Meta Learning
KW - Unreliable Labels
UR - http://www.scopus.com/inward/record.url?scp=85200751463&partnerID=8YFLogxK
U2 - 10.1109/YAC63405.2024.10598519
DO - 10.1109/YAC63405.2024.10598519
M3 - Conference contribution
AN - SCOPUS:85200751463
T3 - Proceedings - 2024 39th Youth Academic Annual Conference of Chinese Association of Automation, YAC 2024
SP - 1702
EP - 1706
BT - Proceedings - 2024 39th Youth Academic Annual Conference of Chinese Association of Automation, YAC 2024
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 39th Youth Academic Annual Conference of Chinese Association of Automation, YAC 2024
Y2 - 7 June 2024 through 9 June 2024
ER -