🤖🔗🧠 Model Based Inference of Synaptic Plasticity Rules

MetaLearnPlasticity analyzes experimental data from the brain and behavior to discover the rules governing how neurons form and modify their connections (synaptic plasticity). It works by creating mathematical models of these plasticity rules and using gradient descent to find the model parameters that best explain how neural connections change during learning.

🌐 Visit paper website (NeurIPS'24)

---

🚀 Features

• Parameterized Plasticity Rules
  Models plasticity using Taylor series expansions or multilayer perceptrons for flexibility and interpretability.

• Gradient Descent Optimization
  Optimizes parameters over entire trajectories to match observed neural or behavioral data.

• Nonlinear Dependencies
  Capable of learning rules with intricate, nonlinear dependencies, such as postsynaptic activity and synaptic weights.

• Simulation Validation
  Validates method by recovering known rules (e.g., Oja's rule) and exploring complex hypothetical rules.

---

🛠 Installation

This project uses Poetry for dependency management. Follow these steps to set up the environment:

1. Install Poetry
   Poetry simplifies dependency management and packaging. Install it with:

   curl -sSL https://install.python-poetry.org | python -

2. Clone the Repository

   git clone https://github.com/yashsmehta/MetaLearnPlasticity.git
   cd MetaLearnPlasticity

3. Install Dependencies
   Run the following command to install all required dependencies:

   poetry install

4. Activate the Virtual Environment
   To activate the project's virtual environment:

   poetry shell

---

📂 Code Structure

• plasticity/: Main project directory containing core modules:

  • run.py: Entry point for experiments; configure and start training here.
  • synapse.py: Handles synaptic plasticity initialization and operations.
  • data_loader.py: Preprocesses data from experiments or simulations.
  • losses.py: Defines custom loss functions for optimization.
  • model.py: Implements the neural network model with parameterized plasticity functions.
  • utils.py: Contains utility functions for logging and data transformation.
  • inputs.py: Manages input stimuli for the model.
  • trainer.py: Encapsulates the training loop, evaluation, and related processes.

• pyproject.toml: Configuration file for Poetry, listing all project dependencies.

---

🧪 Getting Started

1. Set up the environment as described in the Installation section.
2. Modify configurations in run.py to customize experiments.
3. Run the script:

   python plasticity/run.py

---

📖 Citation

If you use this work, please cite it as follows:

@inproceedings{metalearn-plasticity-2024,
  title={Model-Based Inference of Synaptic Plasticity Rules},
  author={Mehta, Yash and Tyulmankov, Danil and Rajgopalan, Adithya and Turner, Glenn and Fitzgerald, James and Funke, Jan},
  booktitle={Advances in Neural Information Processing Systems (NeurIPS) 2024},
  year={2024},
  url={https://neurips.cc/},
}

---

🌟 Contributing

Contributions are welcome! Feel free to submit issues or pull requests. For significant changes, please discuss them via an issue first.

---

📜 License

This project is licensed under the MIT License.