ICLR 2026 · Conference paper

Discovering heterogeneous synaptic plasticity rules via large-scale neural evolution

Ziyuan Ye, Beichen Huang, Yujie Wu, Guozhang Chen, Jibin Wu

Large-scale multi-objective evolution in a realistic mouse visual cortex model uncovers biologically plausible, heterogeneous synaptic plasticity rules.

The scientific question

How do diverse synaptic plasticity mechanisms coordinate to produce functional learning?

Cortical synapses use diverse cell-type-specific plasticity rules, but how heterogeneous local mechanisms jointly give rise to adaptive behavior remains unclear.

Framework for discovering heterogeneous synaptic plasticity rules.
Figure 1 from the original paper ↗

The core idea

Multi-Objective Evolutionary Rule Discovery

We parameterize pre- and postsynaptic spikes, eligibility traces, and neuromodulation via Taylor expansions inside a mouse V1 model, evolving over 2,600 plasticity parameters against biological and task objectives.

What we found

Computational Degeneracy and Few-Shot Learning

Evolution discovers mathematically distinct plasticity families that achieve comparable visual performance, demonstrating computational degeneracy alongside cross-scale robustness and rapid few-shot adaptation.

Continue the question

Related reading

Explore our research

Cite this work

Ziyuan Ye, Beichen Huang, Yujie Wu, Guozhang Chen, Jibin Wu. Discovering heterogeneous synaptic plasticity rules via large-scale neural evolution. ICLR 2026.