New open-source CANNs toolkit unifies brain-inspired attractor neural network research

Continuous attractor neural networks (CANNs) are the leading computational framework for understanding how the brain encodes continuous variables—spatial position, head direction, movement direction—explaining the activity of place cells, grid cells, and head-direction cells. Despite their importance, CANN research has been fragmented: labs rely on custom implementations, general-purpose simulators lack CANN-specific abstractions, and analyzing real neural recordings for attractor geometry has no standardized pipeline. A new toolkit from Sichao He and six co-authors aims to solve this.

CANN is a comprehensive open-source suite with three integrated components. The first is `canns`, a Python library built on BrainPy/JAX that offers standardized 1D and 2D CANNs, spike-frequency-adaptation variants, grid cell networks, hierarchical path-integration models, and brain-inspired attractor architectures—plus curated datasets, task generators, and trainer modules for biologically plausible plasticity. The second is `canns-lib`, a Rust acceleration backend that delivers hundreds-of-times speedups for spatial-navigation workloads and modest gains for Ripser-based persistent homology computations. The third is ASA (Attractor Structure Analyzer), a PySide6 pipeline that applies persistent homology and cohomology to experimental neural recordings to detect ring-like and toroidal attractor signatures.

The toolkit ships with fully reproducible pipelines that recover recent CANN results, including SFA-driven anticipative tracking, theta sweeps in head-direction/place/grid systems, and hierarchical path integration. By standardizing the full research workflow—from model building to recording analysis—CANN removes barriers to entry and accelerates progress in understanding neural representations of space and movement.