Examples

The repository ships runnable scripts under examples/. They are meant to be small, focused walkthroughs of the main API patterns. The descriptions below explain what each script is meant to teach so users can jump straight to the example that matches their workflow.

Core Examples

Script	Use case
example_single_trial_single_channel.py	Smallest forward-model example: one predictor, one output, one trial
example_multi_trial_single_channel.py	Multi-trial fit with cross-validation and explicit trial weighting
example_multifeature_multichannel.py	Multiple predictors and multiple outputs, including grid plotting
example_banded_regularization.py	Grouped predictor regularization for feature blocks with different ridge penalties
example_multitaper_estimator.py	DPSS multi-taper estimation through train_multitaper(...)
example_frequency_resolved_weights.py	Frequency-resolved lag-domain maps and spectrogram-like kernel views
example_backward_decoding.py	Backward decoding with direction=-1
example_bootstrap_confidence_interval.py	Stored bootstrap intervals and uncertainty-aware kernel plots
example_trial_weighting.py	Inverse-variance trial weighting and weighted vs unweighted fits
example_save_and_load.py	Serialization, deserialization, and impulse-response export

Comparison and Benchmarking

Script	Use case
compare_with_mtrf.py	Synthetic kernel comparison against time-domain references
example_mtrf_sample_eeg.py	Public speech-EEG comparison against mTRF, with neg_mse lambda selection and held-out Pearson reporting for a forward benchmark plus a backward compressed-envelope benchmark that reconstructs a p=0.4 broadband target using segmented Hann windows and a wider lambda search in ffTRF
benchmark_runtime.py	Runtime benchmark against mTRF under several scenarios

Which Example Should I Start With?

• Start with example_single_trial_single_channel.py if you want the shortest possible end-to-end script.
• Start with example_multi_trial_single_channel.py if your real data come in multiple trials and you expect to use cross-validation.
• Start with example_multifeature_multichannel.py if your predictors are multi-dimensional or your response has several channels.
• Start with example_backward_decoding.py if your use case is decoding rather than forward encoding.
• Start with example_multitaper_estimator.py if you already know you want the DPSS workflow.

Running Examples

Core examples:

python examples/example_single_trial_single_channel.py
python examples/example_multi_trial_single_channel.py
python examples/example_multitaper_estimator.py

Optional comparison environment:

pixi run -e compare compare-demo
pixi run -e compare benchmark-demo
pixi run -e compare python examples/example_mtrf_sample_eeg.py

Rendered Notebooks

If you want a more tutorial-style presentation than the plain scripts, the docs site also renders lightweight notebooks:

• Getting Started Notebook
• Frequency-Resolved Notebook

These notebooks mirror the same public API as the scripts while interleaving code, explanation, and representative plots.

What to Look For

When reading the examples, pay attention to:

• how single arrays differ from lists of trials
• how direction changes which side is treated as predictor vs target
• when train(...) returns None versus cross-validation scores
• how the same fitted model can be inspected with lag-domain plots, frequency-domain diagnostics, and bootstrap intervals

Gallery