MultiDiffNet: A Multi-Objective Diffusion Framework for Generalizable Brain Decoding

2025-11-23

AI Summary

EEG-based neural decoding suffers from poor cross-subject generalization due to high inter-subject variability and scarcity of large-scale labeled data. To address this, we propose MultiDiffNet—a novel framework that departs from generative data augmentation and instead introduces the first diffusion-based multi-objective joint optimization for EEG. It unifies modeling of cross-task (SSVEP, motor imagery, P300, imagined speech) and cross-session neural features within a compact latent space. We establish the first EEG-specific, multi-task unified benchmark and design a statistical evaluation protocol tailored to low-trial regimes. Extensive experiments demonstrate that MultiDiffNet achieves state-of-the-art cross-subject generalization across all four canonical paradigms, significantly enhancing the practicality and reproducibility of brain–computer interface systems.

Technology Category

Cognitive Modeling & Cognitive Systems: Neural Spike CodingMachine Learning: Multimodal LearningComputer Vision: Diffusion Models for Vision

Application Category

Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasetsResponsible Web: Machine-in-the-loop, human agency and autonomyWeb Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

Neural decoding from electroencephalography (EEG) remains fundamentally limited by poor generalization to unseen subjects, driven by high inter-subject variability and the lack of large-scale datasets to model it effectively. Existing methods often rely on synthetic subject generation or simplistic data augmentation, but these strategies fail to scale or generalize reliably. We introduce extit{MultiDiffNet}, a diffusion-based framework that bypasses generative augmentation entirely by learning a compact latent space optimized for multiple objectives. We decode directly from this space and achieve state-of-the-art generalization across various neural decoding tasks using subject and session disjoint evaluation. We also curate and release a unified benchmark suite spanning four EEG decoding tasks of increasing complexity (SSVEP, Motor Imagery, P300, and Imagined Speech) and an evaluation protocol that addresses inconsistent split practices in prior EEG research. Finally, we develop a statistical reporting framework tailored for low-trial EEG settings. Our work provides a reproducible and open-source foundation for subject-agnostic EEG decoding in real-world BCI systems.

Problem

• Addresses poor EEG generalization across subjects due to variability
• Overcomes limitations of synthetic data generation and augmentation
• Solves inconsistent evaluation practices in EEG decoding research

Innovation

• Diffusion framework learns compact latent space
• Decodes directly from multi-objective latent space
• Achieves state-of-the-art cross-subject generalization