DECODE: Dual-Enhanced COnditioned Diffusion for EEG Forecasting
Mehran Shabanpour, Sadaf Khademi, Konstantinos N Plataniotis, Arash Mohammadi · 2026
Read DECODE: Dual-Enhanced COnditioned Diffusion for EEG Forecasting by Mehran Shabanpour, Sadaf Khademi, Konstantinos N Plataniotis, Arash Mohammadi (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-16885.txt.
Introduction
Shabanpour and co-authors present DECODE, a diffusion framework that forecasts EEG signals during thinking tasks. Brain signals are hard to forecast because they are noisy and shaped by context. DECODE guides the diffusion process with plain language descriptions of each event. It also uses past signals to keep the forecast coherent over time. On a driving dataset with five behaviors, the authors report a mean error of 0.626 microvolts across 75 steps. Its uncertainty estimates stayed well calibrated. The design also points toward use on new behaviors with no retraining.
Catalog record
| CREATOR | Mehran Shabanpour, Sadaf Khademi, Konstantinos N Plataniotis, Arash Mohammadi |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 24304 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.16885
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
DECODE: Dual-Enhanced COnditioned Diffusion for EEG Forecasting Mehran Shabanpour, Sadaf Khademi, Konstantinos N Plataniotis, Arash Mohammadi Abstract Forecasting Electroncephalography (EEG) signals during cognitive events remains a fundamental challenge in neuroscience and Brain-Computer Interfaces (BCIs), as existing methods struggle to capture both the stochastic nature of neural dynamics and the semantic context of behavioral tasks. We present the Dual-Enhanced COnditioned Diffusion (DECODE) for EEG, a novel framework that unifies semantic guidance from natural language descriptions with temporal dynamics from historical signals to generate event-specific neural responses. DECODE leverages pre-trained language models to condition the diffusion process on rich textual descriptions of cognitive events, while maintaining temporal coherence through history-based Langevin dynamics. Evaluated on a real-world driving task dataset with five distinct behaviors, DECODE achieves sub-microvolt prediction accuracy (MAE = 0.626 μ V =~0.626~\mu\text{V} ) over 75-timestep horizons while maintaining well-calibrated uncertainty estimates. Our framework demonstrates that natural language can effectively bridge high-level cognitive descriptions and low-level neural dynamics, opening new possibilities for zero-shot generalization to novel behaviors and interpretable BCIs. By generating physiologically plausible, event-specific EEG trajectories conditioned on semantic descriptions, DECODE establishes a new paradigm for understanding and predicting context-dependent neural activity. † † address: † Concordia Institute for Information Systems Engineering (CIISE), Concordia University, Canada. ‡ Department of Electrical and Computer Engineering, University of Toronto, Canada 1 Introduction Figure 1: Architecture of the proposed dual-conditioned diffusion model for event-specific EEG forecasting. The DECODE framework integrates semantic guidance from natural language event descriptions (top path) with history-based temporal conditioning (bottom path) through a hierarchical diffusion process. Text descriptions are encoded via frozen BERT-large and projected to a shared embedding space, while historical EEG signals undergo forward diffusion before processing through encoder-decoder transformers with interpretable decomposition. The ability to accurately forecast future brain patterns based on historical neural signals has become increasingly critical for applications ranging from seizure prediction to real-time cognitive load assessment and brain-state-dependent stimulation protocols [ 5 ] . In this context, Electroencephalography (EEG) signals provide a window into the complex dynamics of brain activity, offering unprecedented opportunities for understanding cognitive states, predicting neural responses, and developing Brain-Computer Interfaces (BCIs). The inherently non-stationary, multi-scale, and subject-specific nature of EEG signals, however, presents fundamental challenges that have limited the effectiveness of traditional forecasting approaches. The complexity of EEG forecasting extends beyond simple time series prediction. Neural signals exhibit intricate spatio-temporal correlations across multiple frequency bands [ 2 , 9 ] , with dynamics that vary across individuals, cognitive states, and environmental contexts. Recent studies have demonstrated that cognitive events and behavioral responses manifest as distinct patterns in EEG recordings, particularly in high-stakes scenarios such as driving tasks where split-second neural responses can determine critical outcomes [ 18 , 1 ] . These observations underscore the need for forecasting models that can capture both the stochastic nature of neural dynamics and the deterministic patterns associated with specific cognitive events. Traditional approaches to EEG forecasting, including autoregressive models and Recurrent Neural Networks (RNNs), have shown limited success in capturing the full distribution of possible future trajectories, particularly when conditioned on complex cognitive events [ 12 ] . While these methods can estimate conditional means effectively, they struggle to model the uncertainty and multi-modal nature of future EEG states, particularly when conditioned on complex cognitive events. This limitation becomes especially pronounced in real-world applications where understanding the range of possible neural responses is as important as predicting the most likely outcome. The paper targets addressing this limitation by capitalizing on recent advances in generative modeling, particularly, diffusion models, which have revolutionized time series forecasting [ 8 , 22 ] . Unlike traditional discriminative approaches, diffusion models learn to reverse a gradual noising process, allowing them to capture the full conditional distribution of future states while maintaining temporal coherence [ 4 , 20 ] . For the targeted task of EEG forecasting, this capability is particularly valuable as it enables the generation of physiologically plausible signals that respect the complex spectral and spatial characteristics of neural activity. Literature Review: Diffusion models for time series forecasting have progressed from early conditional approaches such as TimeGrad [ 8 ] , which suffered from error accumulation and limited long-range dependencies, to self-supervised masking strategies such as CSDI and SSSD, enabling parallel generation [ 21 , 22 ] . More recent works have focused on enhancing conditioning and representation; TimeDiff [ 15 ] addresses boundary issues, TMDM [ 8 ] leverages transformer architectures capturing uncertainty in multivariate forecasting tasks, Diffusion-TS [ 22 ] preserves trend and seasonality via Fourier-based losses, CN-Diff [ 14 ] integrates nonlinear time-dependent transformations to capture complex temporal patterns, and D3U [ 7 ] separates deterministic and uncertain components for improved probabilistic forecasting. Despite these advances, existing methods primarily rely on historical observations or simple categorical conditions, lacking the flexibility to incorporate rich semantic information about future events. When it comes to EEG forecasting, traditional models relied on linear autoregressive models, which remain practical but perform poorly for long-range predictions. Recent deep learning methods, such as WaveNet [ 12 ] , enhance conditional probability modeling of future samples, particularly for theta and alpha bands. Multi-channel EEG forecasting, despite its importance for capturing whole-brain dynamics, has received limited attention. EEGDIF [ 5 ] addressed this gap with a diffusion model achieving correlation coefficients above 0.77 across 16 channels. However, most approaches remain limited to resting-state or seizure EEG, with little focus on event-conditioned forecasting in complex cognitive tasks. Application of diffusion models to EEG data presents unique challenges due to the signals’ complex spatio-temporal structure and subject-specific variations. EEGDfus [ 4 ] introduced a dual-branch CNN-Transformer architecture for denoising, highlighting the importance of multi-scale feature extraction. DiffEEG [ 16 ] leveraged diffusion-based data augmentation conditioned on short-time Fourier spectrograms to address class imbalance in seizure prediction. Other approaches focus on EEG reconstruction and synthesis, with STAD [ 20 ] using multi-scale Transformers for super-resolution reconstruction, and classifier-free guidance generating subject-, session-, and class-specific Event-Related Potentials (ERPs) [ 6 ] . Understanding EEG dynamics during cognitive tasks provides crucial context for developing event-specific forecasting models. EEG studies in driving tasks have identified distinct neural signatures associated with different driving behaviors. In particular, beta-band Event-Related Desynchronization/Synchronization (ERD/ERS) and fronto-parietal activity reflect motor preparation and cognitive load during driving, highlighting their relevance for modeling task-specific neural dynamics into forecasting models [ 18 , 1 ] . Contributions: We propose a novel conditional diffusion framework for event-specific EEG forecasting that incorporates semantic information to guide signal generation, enabling preemptive safety interventions in semi-autonomous vehicles. Our key insight is that natural language descriptions of events contain rich contextual information that can guide the diffusion process toward generating event-appropriate neural responses. This intuition goes beyond discrete event labels, capturing the richness of cognitive events [ 18 ] . In summary, the paper makes the following contributions: • The Dual-Enhanced COnditioned Diffusion (DECODE) framework is introduced, which is the first diffusion-based EEG forecasting model that simultaneously leverages natural language descriptions for semantic guidance and historical signals for temporal coherence. The proposed dual conditioning mechanism addresses the critical challenge of latent space collapse when generating signals with subtle inter-class differences. • Introduction of a Semantic-Neural Bridge (SNB) within the underlying