Spiking RNN

The spiking package provides leaky integrate-and-fire (LIF) spiking neural networks mapped from continuous rate RNNs.

Core Modules

• LIF Network Function
  • Main Function
  • Parameters
  • Returns
  • Example Usage
• Go-NoGo Evaluation
  • Main Function
  • Function Parameters
  • Example Usage
  • Output Metrics
  • Visualization
• Scaling Factor Grid Search
  • Main Functions
  • Grid Search Parameters
  • Single Trial Evaluation
  • Example Usage
  • Optimization Process
  • Performance Metrics
  • Output Files
  • Output
• Utilities
  • Model Loading
  • Network Configuration
  • Validation Functions
  • Analysis and Formatting
  • Example Usage
  • Function Details

Module Overview

LIF_network_fnc.py

Core function for converting rate RNNs to spiking networks and running LIF simulations.

eval_go_nogo.py

Evaluation functions for testing spiking network performance on Go-NoGo tasks.

lambda_grid_search.py

Grid search optimization for finding optimal scaling factors in rate-to-spike conversion.

utils.py

Utility functions for model loading, network configuration, and spike data analysis.

Quick Reference

Main Functions:

• LIF_network_fnc(): Core rate-to-spike conversion and simulation

• eval_go_nogo(): Evaluate Go-NoGo task performance

• lambda_grid_search(): Optimize scaling factors

Configuration:

• SpikingConfig: Configuration dataclass for spiking RNN parameters

• create_default_spiking_config(): Create default configuration

Utility Functions:

• load_rate_model(): Load PyTorch rate models

• format_spike_data(): Format spike data for analysis