Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Backpropagation-Free Test-Time Adaptation for Lightweight EEG-Based Brain-Computer Interfaces

Siyang Li, Jiayi Ouyang, Zhenyao Cui, Ziwei Wang, Tianwang Jia, Feng Wan, Dongrui Wu

TL;DR

The paper addresses inter-subject variability and nonstationarity in EEG-based BCIs and the limitations of gradient-based TTA on resource-constrained devices. It presents Backpropagation-Free Transformations (BFT), which generates multiple test-time representations via knowledge-guided augmentations (BFT-A) and approximate Bayesian inference (BFT-D), weighting them with a learned ranking module to refine predictions without backprop. A variance-based theoretical foundation justifies why aggregating diverse, potentially correlated branches reduces uncertainty under domain shifts, with formulas such as $\mathrm{Var}(\hat{f}_{\mathbf{w}}) = \sum_{k=1}^{K} w_k^2 \mathrm{Var}(f_k) + \sum_{i \neq j} w_i w_j \mathrm{Cov}(f_i, f_j)$. Empirically, BFT improves MI classification and driver-drowsiness regression across five EEG datasets, remains robust under test-time noise, and remains compatible with quantization for edge deployment, enabling plug-and-play BCIs on resource-constrained devices.

Abstract

Electroencephalogram (EEG)-based brain-computer interfaces (BCIs) face significant deployment challenges due to inter-subject variability, signal non-stationarity, and computational constraints. While test-time adaptation (TTA) mitigates distribution shifts under online data streams without per-use calibration sessions, existing TTA approaches heavily rely on explicitly defined loss objectives that require backpropagation for updating model parameters, which incurs computational overhead, privacy risks, and sensitivity to noisy data streams. This paper proposes Backpropagation-Free Transformations (BFT), a TTA approach for EEG decoding that eliminates such issues. BFT applies multiple sample-wise transformations of knowledge-guided augmentations or approximate Bayesian inference to each test trial, generating multiple prediction scores for a single test sample. A learning-to-rank module enhances the weighting of these predictions, enabling robust aggregation for uncertainty suppression during inference under theoretical justifications. Extensive experiments on five EEG datasets of motor imagery classification and driver drowsiness regression tasks demonstrate the effectiveness, versatility, robustness, and efficiency of BFT. This research enables lightweight plug-and-play BCIs on resource-constrained devices, broadening the real-world deployment of decoding algorithms for EEG-based BCI.

Backpropagation-Free Test-Time Adaptation for Lightweight EEG-Based Brain-Computer Interfaces

TL;DR

The paper addresses inter-subject variability and nonstationarity in EEG-based BCIs and the limitations of gradient-based TTA on resource-constrained devices. It presents Backpropagation-Free Transformations (BFT), which generates multiple test-time representations via knowledge-guided augmentations (BFT-A) and approximate Bayesian inference (BFT-D), weighting them with a learned ranking module to refine predictions without backprop. A variance-based theoretical foundation justifies why aggregating diverse, potentially correlated branches reduces uncertainty under domain shifts, with formulas such as . Empirically, BFT improves MI classification and driver-drowsiness regression across five EEG datasets, remains robust under test-time noise, and remains compatible with quantization for edge deployment, enabling plug-and-play BCIs on resource-constrained devices.

Abstract

Electroencephalogram (EEG)-based brain-computer interfaces (BCIs) face significant deployment challenges due to inter-subject variability, signal non-stationarity, and computational constraints. While test-time adaptation (TTA) mitigates distribution shifts under online data streams without per-use calibration sessions, existing TTA approaches heavily rely on explicitly defined loss objectives that require backpropagation for updating model parameters, which incurs computational overhead, privacy risks, and sensitivity to noisy data streams. This paper proposes Backpropagation-Free Transformations (BFT), a TTA approach for EEG decoding that eliminates such issues. BFT applies multiple sample-wise transformations of knowledge-guided augmentations or approximate Bayesian inference to each test trial, generating multiple prediction scores for a single test sample. A learning-to-rank module enhances the weighting of these predictions, enabling robust aggregation for uncertainty suppression during inference under theoretical justifications. Extensive experiments on five EEG datasets of motor imagery classification and driver drowsiness regression tasks demonstrate the effectiveness, versatility, robustness, and efficiency of BFT. This research enables lightweight plug-and-play BCIs on resource-constrained devices, broadening the real-world deployment of decoding algorithms for EEG-based BCI.
Paper Structure (26 sections, 3 theorems, 26 equations, 8 figures, 6 tables, 1 algorithm)

This paper contains 26 sections, 3 theorems, 26 equations, 8 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Works
  3. Transfer Learning
  4. Test-Time Adaptation
  5. Test-Time Adaptation Beyond Model Update
  6. Backpropagation-Free Transformations for Test-Time Adaptation
  7. Problem Formulation
  8. Test-Time Transformations
  9. Learning-to-Rank Transformations
  10. Inference Aggregation
  11. Summary of BFT
  12. Theoretical Foundation for BFT
  13. Label-Preserving Test-Time Randomization
  14. Variance Decomposition for Weighted Aggregation
  15. Homogeneous Variance Case
  16. ...and 11 more sections

Key Result

Lemma 1

Fix an input $\mathbf{x}$. Let $f_k:=f(\zeta_k;\mathbf{x})$ be square-integrable random variables induced by the joint test-time randomness, and define $\mu_k:=\mathbb{E}[f_k]$ for $k=1,\dots,K$. Let $\hat{f}_{\mathbf{w}}:=\sum_{k=1}^K w_k f_k$ with deterministic weights $\mathbf{w}$. Then we obtain

Figures (8)

  • Figure 1: Key issues in deploying TTA algorithms for BCI decoding.
  • Figure 2: Two types of transformations. (a) BFT-A; and (b) BFT-D.
  • Figure 3: Training and inference of the ranking module, and prediction aggregation strategy for classification and regression tasks, respectively.
  • Figure 4: Illustration of uncertainty reduction achieved through test-time transformations under the homogeneous variance assumption.
  • Figure 5: Two types of test-time noise, using an EEG trial from Zhou2016 as an example. (a) temporal noise; and (b) spatial noise.
  • ...and 3 more figures

Theorems & Definitions (4)

  • Definition 1: Test-Time Randomization and Aggregation
  • Lemma 1: Exact Variance Decomposition
  • Theorem 1: Uncertainty Reduction under Homogeneous Variance
  • Theorem 2: Uncertainty Reduction under Heterogeneous Variance