Bayesian and Frequentist EEG Classification: Model Scores and Compute Profiling

Ethan Davis published model performance scores, compute profiling metrics, and MCMC convergence diagnostics on 2026-05-05. The data originates from a large-scale benchmark comparing Bayesian and frequentist pipelines for motor imagery EEG classification across twenty publicly available MOABB datasets and six pipeline pairs. Performance was assessed using six metrics including AUROC and MCC, with compute data covering training times.

Use Cases

• Benchmarking EEG classification pipelines based on reported performance metrics like AUROC and MCC
• Analyzing MCMC convergence for Bayesian models based on diagnostics including R-hat and effective sample size
• Profiling computational costs of machine learning models based on training time data across datasets
• Comparing classical feature extraction (CSP, Tangent Space) and deep learning architectures (ShallowConvNet, DeepConvNet) for EEG

Strengths

• Includes six distinct performance metrics (AUROC, MCC, NLL, Brier score, ECE, MCE) for model evaluation
• Benchmark spans twenty publicly available MOABB datasets, providing a broad evaluation base
• Contains multiple MCMC convergence diagnostics (R-hat, effective sample size, divergences, visual plots)
• Data covers six different machine learning pipeline pairs for comparison

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• Description metadata is limited; actual data quality requires manual inspection after download

Provenance

Source

figshare

Collection Method

Results from a large-scale benchmark study comparing machine learning pipelines

Freshness

Last updated 2026-05-05 17:29:16; freshness should be verified