Loading
Life sciences · Preprint
arXiv · September 10, 2026
Early or partial results. Treat as a signal, not a conclusion.
This preprint describes a spatial-partitioning variant of ES-HyperNEAT that mitigates a central-bias failure mode on MNIST, achieving 43% mean accuracy versus 21% baseline—a 106% relative improvement. The gain stems from forcing evolved networks to discover features across the entire image (expanding pixel coverage from 4% to 79%), not from weighting aggregation; however, absolute accuracy remains substantially below gradient-trained methods, and the work is limited to a single toy benchmark.
Computational diagnostic and methods development study. MNIST digit classification benchmark; no empirical population studied.. Intervention: 13-expert spatially partitioned ES-HyperNEAT architecture with equal-weighted aggregation. Compared with: Baseline ES-HyperNEAT (single network); gradient-trained methods mentioned as external reference but no direct comparison reported.
Baseline ES-HyperNEAT achieves 21% mean accuracy on MNIST due to spatial-concentration bias at image centre 13-expert spatially partitioned design reaches 43% mean accuracy, a 106% relative improvement over baseline Equal-weighted averaging of experts yields 70% improvement without validation-data-driven aggregation
Safety was not reported in the material analysed. Check the source before drawing any conclusion about harm.
The source did not state who this applies to in practice.
This is an early-stage computational study demonstrating a workaround for a known failure mode in neuroevolution on a toy benchmark (MNIST), with no comparison to established neuroevolution baselines and results still well below gradient-trained methods.
As stated by the source record.
Quoted from the source exactly as published.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
Evolvable-Substrate HyperNEAT (ES-HyperNEAT), a bio-inspired indirect encoding that determines neuron placement and connection weights from spatial coordinates, exhibits a failure mode on MNIST as a diagnostic benchmark. Because input pixels map to a coordinate space centered at the origin, evolved networks converge on a small central cluster of input pixels, a spatial-concentration bias; prior work observed only 21% mean accuracy in this regime. Is this bias an optimization artifact or an architectural ceiling? Inspired by Mixture-of-Experts (MoE) principles, we partition the input into non-overlapping spatial segments, each assigned to a separately evolved specialist network. With 13 such experts, this design reaches 43% mean accuracy, a 106% relative improvement over the baseline. The architectural gain does not depend on data-driven aggregation: equal-weighted averaging, which uses no validation data, already yields a 70% improvement; the gain comes from partitioning, not the weighting. Receptive-field analysis shows the mechanism: partitioning forces evolution to discover features across the entire image, expanding active pixel coverage from 4% to 79%. Absolute accuracy stays below gradient-trained baselines, but the relative gain points to central bias, not the evolutionary search. Two tools are designed to generalize beyond MNIST: a receptive-field diagnostic for silent input-coverage collapse, and a spatial-partitioning remedy that restores coverage.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.