Adaptive optimizer improves eeg decoding across different people
Adaptive Forgetting for Nonstationary Optimization: Towards Robust EEG Decoding
Machine LearningHuman-Computer Interaction
Summary
It is hard for brain signal decoding systems to perform well when used on different people because brain signals vary and are noisy. The authors propose a new method called AFOR that adjusts how it learns by watching the signals' behavior and changing its settings on the fly. This method helps the system adapt better to the differences between individuals and improves its accuracy in recognizing brain activity patterns. Tests on several EEG datasets showed that AFOR outperformed standard methods by a few percentage points.
What this means in practice
- •For neurotechnology developers: Create brain-computer interfaces that maintain accuracy when used by different individuals by using adaptive optimization for model training.
- •For medical device engineers: Improve EEG-based diagnostic tools by enhancing their robustness across patient variability through dynamic parameter adjustment in learning algorithms.
Authors
Hongyu Zhu, Lin Chen, Jing Chen, Yuting Zhou, Mingsheng Shang
Abstract
Electroencephalography (EEG) provides non-invasive monitoring of brain activity and is widely used in emotion recognition, motor imagery and sleep staging. Although within-subject decoding has achieved considerable progress, cross-subject generalization remains a central challenge in practical applications. EEG decoders are typically trained with Adam/AdamW under a fixed second-moment decay coefficient, even though cross-subject learning involves low signal-to-noise ratios, subject variability, and gradient nonstationarity. A fixed coefficient implicitly assumes that gradient statistics are homogeneous across layers and time, which can limit model's adaptability to cross-subject EEG signals and degrade generalization. To address these issues, we propose AFOR, a tensor-wise adaptive optimizer that converts the fixed second-moment decay coefficient into a dynamic coefficient estimated online from local gradient state. AFOR combines a Residual-Alignment Signal Scorer (RASS) and an Adaptive Forgetting Controller (AFC). RASS summarizes local gradient residuals and directional agreement into a signal-quality score, and AFC maps this score through self-referential normalization to a bounded per-step decay coefficient, with cumulative-product initialization correction maintaining consistency under time-varying decay. Under a strict cross-subject protocol on three EEG benchmarks that cover three representative fields, AFOR achieves the best average performance among the compared optimizers, improving the mean test accuracy over Adam by 3.00%, 2.07%, and 4.38%, respectively.