Saved in:
Bibliographic Details
Main Authors: Medvid, Sergii, Valenia, Andrii, Glybovets, Mykola
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.03386
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917387429740544
author Medvid, Sergii
Valenia, Andrii
Glybovets, Mykola
author_facet Medvid, Sergii
Valenia, Andrii
Glybovets, Mykola
contents Developmental approaches to neural architecture search grow functional networks from compact genomes through self-organisation, but the resulting networks operate with fixed post-growth weights. We characterise Hebbian and anti-Hebbian plasticity across 50,000 morphogenetically grown recurrent controllers (5M+ configurations on CartPole and Acrobot), then test whether co-evolutionary experiments -- where plasticity parameters are encoded in the genome and evolved alongside the developmental architecture -- recover these patterns independently. Our characterisation reveals that (1) anti-Hebbian plasticity significantly outperforms Hebbian for competent networks (Cohen's d = 0.53-0.64), (2) regret (fraction of oracle improvement lost under the best fixed setting) reaches 52-100%, and (3) plasticity's role shifts from fine-tuning to genuine adaptation under non-stationarity. Co-evolution independently discovers these patterns: on CartPole, 70% of runs evolve anti-Hebbian plasticity (p = 0.043); on Acrobot, evolution finds near-zero eta with mixed signs -- exactly matching the characterisation. A random-RNN control shows that anti-Hebbian dominance is generic to small recurrent networks, but the degree of topology-dependence is developmental-specific: regret is 2-6x higher for morphogenetically grown networks than for random graphs with matched topology statistics.
format Preprint
id arxiv_https___arxiv_org_abs_2604_03386
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Activity-Dependent Plasticity in Morphogenetically-Grown Recurrent Networks
Medvid, Sergii
Valenia, Andrii
Glybovets, Mykola
Robotics
Neural and Evolutionary Computing
Developmental approaches to neural architecture search grow functional networks from compact genomes through self-organisation, but the resulting networks operate with fixed post-growth weights. We characterise Hebbian and anti-Hebbian plasticity across 50,000 morphogenetically grown recurrent controllers (5M+ configurations on CartPole and Acrobot), then test whether co-evolutionary experiments -- where plasticity parameters are encoded in the genome and evolved alongside the developmental architecture -- recover these patterns independently. Our characterisation reveals that (1) anti-Hebbian plasticity significantly outperforms Hebbian for competent networks (Cohen's d = 0.53-0.64), (2) regret (fraction of oracle improvement lost under the best fixed setting) reaches 52-100%, and (3) plasticity's role shifts from fine-tuning to genuine adaptation under non-stationarity. Co-evolution independently discovers these patterns: on CartPole, 70% of runs evolve anti-Hebbian plasticity (p = 0.043); on Acrobot, evolution finds near-zero eta with mixed signs -- exactly matching the characterisation. A random-RNN control shows that anti-Hebbian dominance is generic to small recurrent networks, but the degree of topology-dependence is developmental-specific: regret is 2-6x higher for morphogenetically grown networks than for random graphs with matched topology statistics.
title Activity-Dependent Plasticity in Morphogenetically-Grown Recurrent Networks
topic Robotics
Neural and Evolutionary Computing
url https://arxiv.org/abs/2604.03386