bcijelly 0.7.1

BCI utilities and models

BCIJelly

Welcome

Welcome to BCIJelly, a Python toolkit for invasive BCI workflows.

BCIJelly provides a unified pipeline for:

• loading datasets
• splitting train/val/test data
• creating and training baseline models
• evaluating model performance
• summarizing experiment outputs

SearchAgent note:

• SearchAgent is currently experimental (testing preview).
• The model/task space exposed by the agent is currently not complete.

Install

Install from PyPI:

pip install -U bcijelly

Install LFADS extras only when needed:

pip install -U "bcijelly[lfads]"

Install SearchAgent extras only when needed:

pip install -U "bcijelly[agent]"

Install POYO extras only when needed:

pip install -U "bcijelly[poyo]"

Install chip deployment extras only when needed:

pip install -U "bcijelly[chip]"

Verify installation:

python -c "import bcijelly; print(bcijelly.__file__)"
python -c "from bcijelly import load_data, summary; print('bcijelly import OK')"

Links to Docs

Primary docs entry:

• docs/General/overview.en.md

Demo docs:

• docs/Demo/demo.en.md
• docs/Demo/complex_scenarios.en.md
• docs/Demo/chip_deploy.en.md

API docs:

• docs/API/load_data.en.md
• docs/API/agent.en.md
• docs/API/model.en.md
• docs/API/model.train.en.md
• docs/API/model.test.en.md
• docs/API/summary.en.md
• docs/API/toChip.en.md

License

BCIJelly is licensed under the MIT License.

• LICENSE

Contributors

• Liyuan Han ([e-mail])

References

Project-level publication references are not finalized yet.

• TODO: add official BCIJelly citation when available.

Related model reference:

• Sedler AR, Pandarinath C. lfads-torch: A modular and extensible implementation of latent factor analysis via dynamical systems. arXiv:2309.01230.