Real, measured evidence for the "more biologically plausible neuron models" future-work item

Open
#1 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
25/100
Issue type
Feature
Clarity
Needs clarification
Activity status
Active

Research direction

Start by reading LIF_unit.vhd and the linked heterogeneous-neuron-ontology experiment to understand the current LIF-only implementation and the reported comparison. The issue does not define a specific model, interface, acceptance test, or completion condition, so the concrete scope would need to be agreed before implementation.

Written by the indexing model from the issue text.

Description

Your paper (arXiv:2506.14138) names integration of more biologically plausible neuron models as future work, and the current implementation is LIF-only (LIF_unit.vhd). I ran a small, independent, open experiment testing whether that's worth prioritizing, and wanted to share the real result in case it's useful.

Setup: four neuron models (LIF, Izhikevich, AdEx, a resonator model), each independently verified against its own literature behavior, tested on four synthetic anomaly-detection tasks designed to isolate what each model type can represent.

Result: a heterogeneous population (all four types, same total neuron count, each with its own calibrated threshold) detected 100% of the four anomaly types. A homogeneous LIF-only population of the same total size detected 50%, completely missing the repetition and frequency cases — not by chance, but because LIF structurally cannot represent either signal shape.

Full code and methodology: https://github.com/tritsystem/heterogeneous-neuron-ontology

Given LIF_unit.vhd is already a separate, modular unit rather than something deeply entangled through the core engine, adding an Izhikevich or AdEx variant alongside it looks architecturally plausible without a full redesign — though I haven't attempted a VHDL implementation myself and don't want to assume that's as simple as it looks from the outside.

Not asking for anything specific — just flagging real, measured evidence that this future-work item may be worth prioritizing.

Dominant language
VHDL
Stars
24
Forks
4
PR merge metrics
No merged PRs in 30d

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Similar issues

More AI Infra & Agents issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.