Skip to content

jaxfne v0.2.10

Release date: May 20, 2026

jaxfne v0.2.10 expands the package from a minimal source-to-field/readout scaffold into a tutorial-ready, multimodal workflow package. This release bundles six months of development across field diagnostics, calibration specification, notebook standardization, and three complete Colab-ready tutorials.

What's new

Field and proxy diagnostics (v0.2.4–v0.2.6)

  • Field/proxy admissibility diagnostics — Explicit metadata for source conservation status, row-normalized proxy kernels, boundary conditions, gauge specifications, and source-balance reporting
  • Calibration specification contracts — Standardized documentation for native Izhikevich, normalized, toy-scale, candidate, and empirical calibration workflows
  • Field/proxy diagnostic crosswalk — Clear distinction between proxy readout paths (current) and reserved physical field-solver paths

Notebook standard (v0.2.7)

  • Colab notebook standard — Enforced structure: pip-install-first cell, version verification, cleared outputs, no private paths
  • Static validation — 17 tests ensuring portability across Colab, nbconvert, and CPU environments
  • Documentation — Notebook standard guide with validation checklist and five-notebook progression table

Colab-ready tutorials (v0.2.8–v0.2.10)

Three multimodal source-to-field/readout tutorials, progressively scaled:

  1. Single-neuron (v0.2.8): Single Izhikevich neuron with all eight readout operators
  2. Notebook: notebooks/01_single_neuron_multimodal.ipynb

  3. Example: examples/03_single_neuron_multimodal_probe.py

  4. Two-neuron E/I (v0.2.9): Minimal recurrent network (1E + 1I) with coupling dynamics

  5. Notebook: notebooks/02_two_neuron_ei_multimodal.ipynb

  6. Example: examples/04_two_neuron_ei_multimodal.py

  7. 100-neuron E/I (v0.2.10): Balanced population (75E + 25I) with population-level analysis

  8. Notebook: notebooks/03_network_100_ei_multimodal.ipynb
  9. Example: examples/05_network_100_ei_multimodal.py

Each tutorial demonstrates: - JAX-native configuration and simulation - All eight proxy readout operators (spikes, voltage, source, LFP proxy, CSD proxy, EEG proxy, MEG proxy, EMM proxy) - Output bundle inspection and validation metadata - Reproducible JSON-safe manifests and metric computation

Documentation improvements

  • Professional public landing page — Architecture-first overview with real APIs and example workflows
  • Material theme with Read the Docs — Modern, searchable documentation site at https://jaxfne.readthedocs.io/
  • Updated guides — Probe operators reference, tensor-field workflows, Jaxley interoperability, calibration specifications, output bundles
  • Expanded tutorials table — Version comparison and progression guidance

Highlights

Feature Status
Field/proxy diagnostics Complete (v0.2.4–v0.2.6)
Calibration specification Complete (v0.2.5)
Notebook standard Complete (v0.2.7)
Colab tutorials Complete: 1-neuron (v0.2.8), 2-neuron (v0.2.9), 100-neuron (v0.2.10)
Read the Docs integration Complete
All eight proxy operators Stable, cross-tutorial
JAX vmap/jit CPU-safe, scaling verified to 100 neurons

Validation

  • Test suite: 609 tests passing (baseline 493 → +116 new)
  • Examples: All five core examples pass (00, 02, 03, 04, 05)
  • Notebooks: Three Colab tutorials, outputs cleared, pip-install-first standard enforced
  • Documentation: MkDocs strict build passes, no broken links
  • Output bundles: JSON-strict, NaN/Inf-free, metadata immutable

Installation

Install from PyPI:

pip install jaxfne==0.2.10

Or upgrade from a prior version:

pip install --upgrade jaxfne

Documentation

Full documentation is available at:

https://jaxfne.readthedocs.io/

Start with: - Installation guide - Quickstart - Single-neuron tutorial

Contributing

Questions, feedback, or contributions? See Contributing or open an issue on GitHub.

Citation

If you use jaxfne in your work, please cite:

@software{jaxfne_v0.2.10,
  title={jaxfne: JAX Field Neural Equations},
  author={H N},
  year={2026},
  url={https://github.com/HNXJ/jaxfne},
  version={0.2.10}
}

Scope notes and extension notess

jaxfne is a computational scaffold for source-to-field/readout workflows. It provides: - Reproducible tensor-field operations on JAX-native pipelines - Simulated proxy readout operators with metadata labeling - Calibration specification contracts for calibration workflows

jaxfne does not provide: - Biological validation or empirical calibration (in v0.2.10) - Physical field solvers (proxy operators only) - Whole-brain simulation - Real-time execution

See Scope and limitations for details.

jaxfne v0.2.10 — Field diagnostics, calibration contracts, notebook standards, and multimodal Colab tutorials.