1BT: One-Block Transformer for EEG Workload Assessment
Researchers introduce 1BT, a One-Block Transformer architecture for EEG-based cognitive workload assessment. The model uses a minimal latent bottleneck with a single cross-attention module and lightweight self-attention to aggregate multi-channel temporal sequences. A study with 11 participants performing three tasks (abstract reasoning, numerical problem-solving, interactive video game) across two workload levels was conducted. Systematic analysis identified the most compact configuration preserving high performance while lowering computational cost.
Key facts
- 1BT is a One-Block Transformer for EEG-based cognitive workload assessment.
- The model uses a minimal latent bottleneck with single cross-attention and lightweight self-attention.
- A study involved 11 participants performing three cognitively diverse tasks.
- Tasks included abstract reasoning, numerical problem-solving, and an interactive video game.
- Continuous EEG recordings were taken across two workload levels.
- Systematic architectural analysis identified the most compact configuration.
- The compact configuration preserves high performance with lower computational cost.
- The research aims to balance representational capacity with computational efficiency.
Entities
—