SPICE Documentation

Welcome to the documentation for SPICE (Sparse and Interpretable Cognitive Equations), a framework for automating scientific practice in cognitive science.

Overview

SPICE is built on two fundamental principles:

1. Task-specific RNN Training: A neural network is trained to predict human behavior and implicitly learn latent cognitive mechanisms.

2. Equation Discovery: Using Sparse Identification of nonlinear Dynamics (SINDy) to derive interpretable mathematical equations from the learned mechanisms.

Quick Links

• Installation Guide
• Quick Start Tutorial
• API Reference
• Tutorials

Key Features

• Scikit-learn compatible estimator interface
• Customizable network architecture for identifying complex cognitive mechanisms
• Participant embeddings for identifying individual differences
• Precoded models for:
  • Simple Rescorla-Wagner
  • Forgetting mechanism
  • Choice perseveration
  • Participant embeddings

Getting Help

If you need help using SPICE, you can:

1. Check the tutorials section
2. Visit our GitHub repository
3. Open an issue on GitHub

License

SPICE is released under the MIT License. See the LICENSE file for more details.