NeuroGym provides a curated collection of neuroscience-specific tasks with a unified interface, designed to simplify the training of neural network models on behavioral paradigms relevant to neuroscience research. It builds upon the Gymnasium (formerly OpenAI Gym) framework, enabling seamless integration with existing machine learning algorithms.

How It Works

NeuroGym inherits from Gymnasium, allowing it to leverage a wide array of established machine learning algorithms. It extends this by incorporating features crucial for neuroscience, such as continuous-time and trial-based task structures. The library also includes modifier functions and wrappers that facilitate easy configuration and adaptation of tasks, such as introducing reaction time elements or enabling simultaneous training on multiple tasks.

Quick Start & Requirements

• Install via pip: pip install neurogym
• For editable install: git clone https://github.com/neurogym/neurogym.git && cd neurogym && pip install -e .
• Requires Python 3.11.
• Psychopy installation is noted as currently not working.
• Official documentation: https://neurogym.org/ (Note: The README links to documentation, but the URL itself is not provided in the text.)

Highlighted Details

• Compatible with most Gymnasium-based packages.
• Includes a Jupyter notebook demonstrating reinforcement learning with Stable-Baselines3.
• Provides a template for creating custom tasks in the Gymnasium format.
• Offers a list of implemented tasks and wrappers for task modification.

Maintenance & Community

• The project acknowledges authors via Zenodo DOI and lists other contributors.
• No specific community channels (Discord/Slack) or roadmap links are provided in the README.

Licensing & Compatibility

• The README does not explicitly state the license type.

Limitations & Caveats

• Psychopy installation is currently non-functional.
• The README does not specify the license, which could impact commercial use or closed-source integration.