Spiking Neural Network Continual Learning Code Index Released: 156 Open-Source Projects Empower Neuromorphic Computing Research

The research team from Johannes Kepler University has released a comprehensive code index in the field of spiking neural network continual learning, which includes 156 open-source projects covering six major research directions (architectural methods, replay mechanisms, regularization techniques, biologically inspired plasticity, etc.) and auxiliary categories. This provides a valuable resource navigation tool for the neuromorphic computing community, lowering the threshold for research reproduction and promoting the development of the field.

As the third generation of neural networks, Spiking Neural Networks (SNNs) have attracted much attention in the neuromorphic computing field due to their biological interpretability and low-power characteristics. However, enabling SNNs to have continual learning capabilities (absorbing new knowledge without forgetting old ones in dynamic environments) is a core challenge. The team from Johannes Kepler University has submitted a review paper to Artificial Intelligence Review and simultaneously released a supporting code index repository, providing an open-source resource navigation tool for the community.

This index selects 156 public code projects from 388 literatures, with 146 hosted on GitHub and the rest distributed on platforms like Zenodo and ModelDB. It is organized using a six-branch classification method: architectural methods, replay methods, regularization methods, biologically inspired plasticity, Hebbian projection (currently empty), and hybrid methods; plus four auxiliary categories: neuromorphic hardware platforms, SNN basic theory, CL basic theory, datasets and benchmark tests.

The index construction adopts a three-stage strategy: 1. GitHub title keyword search (targeting each paper's title and method name); 2. Targeted search on alternative platforms (such as Zenodo); 3. Special scanning of top conference proceedings. Quality control mechanisms include: author matching principle, accessibility verification (verify_links.py script), and metadata integrity (including paper title, authors, etc.).

The repository provides a batch cloning script clone_all.sh (supports full/shallow cloning) and a link health check script verify_links.py (checks URL validity). Data formats include CSV (manifest.csv) and JSON (manifest.json, etc.); the index data is licensed under CC BY 4.0, the scripts under MIT license, and third-party code copyright belongs to the original authors.

1. Lowering the reproduction threshold: Concentrating 156 reproducible projects to solve the problem of difficult reproduction in the SNN field; 2. Promoting method comparison: The six-branch classification provides a clear framework for comparing methods in the same category; 3. Supporting hardware deployment: Including code for mainstream neuromorphic chips like Loihi; 4. Promoting standardization: Unifying dataset and benchmark test categories to reduce performance comparison biases.

Limitations: Risk of link decay (it is recommended to run the verification script regularly), possible mismatches in heuristic matching (manual spot checks are recommended), and non-complete mirroring (need to access upstream repositories). Usage recommendations: Verify links before citation, manually check matching status, and comply with the original authors' license agreements.

This index marks that the field of SNN continual learning has entered a more open and reproducible stage. With the commercialization of neuromorphic hardware (such as Loihi2), the demand for open-source code is growing. The team welcomes community contributions (submitted via GitHub Issues). This index provides a solid starting point for scholars and engineers in fields like neuromorphic computing, accelerating the transformation from algorithms to applications.