Teacher-guided multi-task learning for integrated plasma monitoring on EAST
Abstract
This paper integrates tokamak plasma disruption prediction, Edge-Localized Mode (ELM) monitoring, Multifaceted Asymmetric Radiation From the Edge (MARFE) detection, and H/L-mode identification into a unified Multi-Task Learning (MTL) model. This integration aims to realize MTL's potential for improving task performance while reducing inference time. However, on datasets constructed from selected EAST discharges recorded between 2015 and 2020, this study found that the traditional MTL method underperformed the corresponding Single-Task Learning (STL) method on some tasks. This performance degradation is the well-known issue of negative transfer. To address this issue, this study proposes the teacher-guided MTL method, using task-specific STL ensembles (teachers) to guide the training of an MTL model (student). The ensemble teachers guide training through soft-target distillation losses for each task of the MTL student model, and increased weighting of each shot's peak-disruptivity window in the disruption prediction label-supervised loss. Test-set results showed that the teacher-guided MTL method outperformed the corresponding STL method in ROC-AUC for all four tasks, eliminating the negative transfer observed with the traditional MTL method. In addition, on the same single GPU or single physical CPU core, the MTL student model can substantially reduce the inference time required to generate all four task outputs compared with the four STL baseline models. This paper offers a promising route to plasma monitoring systems that cover a broader range of phenomena, improve per-task performance and reduce inference time with limited real-time computational resources, supporting future high-performance plasma operation.
Related
- Machine Learning Accelerated Plasma Simulation through Physics Guided Time Jumps
- Physics is the Best Teacher: Consistency Learning for Time-Invariant Operators of Chaotic Dynamics
- Deflection monitoring of discontinuous built-up beams using FBG measurements with mechanics-guided regularization
- Wavefront-Guided Electron Injection for Direct Laser Acceleration in Relativistic Laser-Driven Plasma Channel
- Development of A STEM-Integrated Guided Inquiry E-Module on Global Warming to Enhance Students’ Conceptual Understanding
BibTeX
@misc{jia2026teacherguided,
title = {{Teacher-guided multi-task learning for integrated plasma monitoring on EAST}},
author = {Yunhu Jia and Wenhui Hu and Bihao Guo and Qingquan Yang and Xin Lin and Haoyu Wang and Yao Huang and Yuehang Wang and Kai Wu and Zhengping Luo and Yanmin Duan and Jilei Hou and Ruirui Zhang and Jianqiu Zhu and Zijie Liu and Qiping Yuan and Bingjia Xiao},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.41134},
eprint = {2610.41134},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.41134},
}
RIS
TY - UNPB
TI - Teacher-guided multi-task learning for integrated plasma monitoring on EAST
AU - Yunhu Jia
AU - Wenhui Hu
AU - Bihao Guo
AU - Qingquan Yang
AU - Xin Lin
AU - Haoyu Wang
AU - Yao Huang
AU - Yuehang Wang
AU - Kai Wu
AU - Zhengping Luo
AU - Yanmin Duan
AU - Jilei Hou
AU - Ruirui Zhang
AU - Jianqiu Zhu
AU - Zijie Liu
AU - Qiping Yuan
AU - Bingjia Xiao
PY - 2026
DA - 2026/10/30
PB - arXiv
DO - 10.48550/arxiv.2610.41134
UR - https://arxiv.org/abs/2610.41134
AB - This paper integrates tokamak plasma disruption prediction, Edge-Localized Mode (ELM) monitoring, Multifaceted Asymmetric Radiation From the Edge (MARFE) detection, and H/L-mode identification into a unified Multi-Task Learning (MTL) model. This integration aims to realize MTL's potential for improving task performance while reducing inference time. However, on datasets constructed from selected EAST discharges recorded between 2015 and 2020, this study found that the traditional MTL method underperformed the corresponding Single-Task Learning (STL) method on some tasks. This performance degradation is the well-known issue of negative transfer. To address this issue, this study proposes the teacher-guided MTL method, using task-specific STL ensembles (teachers) to guide the training of an MTL model (student). The ensemble teachers guide training through soft-target distillation losses for each task of the MTL student model, and increased weighting of each shot's peak-disruptivity window in the disruption prediction label-supervised loss. Test-set results showed that the teacher-guided MTL method outperformed the corresponding STL method in ROC-AUC for all four tasks, eliminating the negative transfer observed with the traditional MTL method. In addition, on the same single GPU or single physical CPU core, the MTL student model can substantially reduce the inference time required to generate all four task outputs compared with the four STL baseline models. This paper offers a promising route to plasma monitoring systems that cover a broader range of phenomena, improve per-task performance and reduce inference time with limited real-time computational resources, supporting future high-performance plasma operation.
ER -