Changelog
Version 1.0
Initial Release
The first public release of the Neulite project.
Updates
bionetlite v1.0.1
Added automatic environment setup for tutorials
bionetlite can be run directly without pre-running BMTK tutorials
Required data files (SWC, JSON, etc.) are automatically downloaded on first run
Tutorial-specific simulation parameters are automatically configured
Fixed parameter order bug in connection.csv
Improved edge filtering
Neulite Kernel v1.0.1
Performance improvement through software pipelining of
solve_matrix()Changed synaptic current processing to implicit method
Fixed bit-shift processing order for delay
Enables large-scale biophysical neuron simulation on the supercomputer Fugaku.
Key Achievements
Large-Scale Simulation Results
Achieved simulation of whole mouse cortex model containing 9 million neurons and 260 billion synapses
Optimized for execution on Fugaku
bionetlite
Network Construction Tool:
NeuliteBuilder class extending BMTK’s NetworkBuilder
Existing BMTK code can be used with only import statement changes
Support for biophysical neuron models
Support for exp2syn synapse model
Automatic preprocessing of morphology files (SWC) - Conversion to Perisomatic model - Conversion to zero-based indexing - Sorting by depth-first search (DFS)
Automatic conversion of ion channel configuration (JSON→CSV)
Support for MPI parallel execution
Neulite
High-Performance Simulator:
Lightweight kernel implemented in C
Specialized in Perisomatic model from Allen Cell Types Database
Support for constant current input
Runs on a wide range of environments from Raspberry Pi to Fugaku
Generated Files
SONATA format (BMTK compatible)
Neulite format -
<network_name>_population.csv- Neuron population definition -<src>_<trg>_connection.csv- Detailed synapse connection information - Processed SWC files - Ion channel configuration CSV -config.h- Simulation configuration header
Documentation
Project Overview
Detailed explanation of bionetlite and Neulite
Setup Guide
4 tutorials - Tutorial 01: Single cell simulation - Tutorial 02: Understanding spike input - Tutorial 03: Single population - Tutorial 04: Multiple populations
User Guide - Basic usage - Configuration files - Parallel execution
Architecture - System overview - Design and implementation (including BMTK and BioNet background)
API Reference - NeuliteBuilder API - File format specification
Advanced Topics - Allen V1 model examples - Specifications and limitations
FAQ
Supported Environments
Linux (recommended)
macOS
Python 3.7 or higher
MPI environment (for parallel execution)
Known Limitations
Supported Models:
Only biophysical models supported (point neuron, virtual neuron not supported)
Only Perisomatic model supported
Supported Synapses:
Only exp2syn supported
Input Methods:
Only constant current input supported
Spike input and Virtual cell input not supported
Others:
Delay values support only int type (decimals are rounded)
Synapse connection positions are randomly determined at network construction time (different from bionet)
Future Plans
Addition of spike input functionality (implementation on Neulite side)
Support for new synapse models
Continuous performance improvement
Documentation expansion
Addition of tutorials