Neural Decoding
Overview
Neural decoding infers what a person perceives, intends, or does from signals such as EEG, MEG, fMRI, ECoG, and neural spikes, supporting neuroscience and brain-computer interfaces. Common approaches include CNNs, Transformers and state-space models, alongside contrastive alignment, generative reconstruction, self-supervised pretraining, and adaptation across subjects or recording sessions; tasks range from speech and image decoding to movement and auditory attention. Researchers evaluate these methods on settings such as the Natural Scenes Dataset, motor-imagery and auditory-attention benchmarks, and speech-decoding datasets, while quantum error-correction decoders use neural networks to infer errors from syndrome measurements. A central challenge is making results reliable beyond the training subject, session, or dataset while controlling for shortcuts, limited data, and real-time latency.
Latest papers
184 papers26 papers in the last four weeks, up 24% on the four weeks before. 0.2% of all new papers.