Reproduce “CANNs: A Toolkit for Research on Continuous Attractor Neural Networks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sichao He, Aiersi Tuerhong, Shangjun She, Tianhao Chu, Yuling Wu, Junfeng Zuo, Si Wu
The CANNs toolkit connects models, task generation, accelerated operations, and analysis of recordings in a shared workflow. Its ASA system uses persistent homology to assess population-activity topology. A screen of 66 medial entorhinal cortex module datasets found heterogeneous topology; selected examples showed ring and toroidal structure, demonstrating analysis pipelines rather than prevalence or proof of an attractor mechanism.
Key insights
- A unified, open-source toolkit effectively addresses the fragmentation problem in computational neuroscience by standardizing implementations and workflows across the entire research lifecycle.
- Integrating a high-performance Rust backend with a flexible Python/JAX frontend substantially enhances computational efficiency, enabling larger-scale simulations and more extensive topological data analyses.
- Topological Data Analysis (TDA), specifically persistent homology and cohomology, offers a robust and interpretable method to infer and characterize underlying attractor geometries (e.g., ring-like, toroidal) from both simulated and experimental neural activity.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.