Determinants of hyperparameter robustness in connectome reservoir computing
Abstract
Reservoir computing provides a controlled setting for studying how recurrent network architectureshapes computation: input signals are projected into a high-dimensional state space by a fixed nonlinear dynamical system, and only the readout is trained. However, reservoir performance can be dependent on hyperparameters; this paper asks which recurrent network features support robustness to those parameter changes. We characterize computational performance using memory capacity (MC), truncated single-delay information-processing capacity (IPC), and kernel rank (KR). Generalization across input histories is measured using generalization rank (GR), while hyperparameter robustness is quantified using the coefficient of variation (CV) of each metric across sweeps of target spectral radius, input scaling, leak rate, and neuron bias. To examine the architectural determinants of robustness, we construct perturbations that alter connectivity topology, excitatory/inhibitory sign structure, weight magnitudes, and weight placement while preserving complementary properties. Across these experiments, the C. elegans connectome consistently occupies a relatively low-variance regime. The central result is a performance-robustness tradeoff: architecture variants with higher task-agnostic performance also tend to exhibit greater hyperparameter sensitivity and poorer common-tail generalization. Across the E/I edge balance sweeps and shuffle controls, this tradeoff is closely associated with the raw spectral radius before normalization. Because every perturbed matrix is rescaled to the same target radius, matrices with lower raw spectral radius receive greater global amplification of their recurrent weights. The observed differences among architectures therefore characterize the joint effects of structural variation and architecture-specific global rescaling under spectral-radius normalization.
Related
- Information coding strategies in biological reservoir computing for multiclass time-series classification
- CLEAR-FORCE: A reservoir computing framework for closed-loop brain stimulation
- Harnessing the Intrinsic Dynamics of Biological Neural Networks for Reservoir Computing
- Learning to play with spikes: characterizing, predicting, and engineering unsupervised plasticity rules for spiking reservoir computing
- Microenvironmental and transcriptional determinants of the robustness of spinal cord regeneration in zebrafish
BibTeX
@misc{churchland2026determinants,
title = {{Determinants of hyperparameter robustness in connectome reservoir computing}},
author = {Miles Walter Churchland and Raul de Palma Aristides and Jordi Garcia-Ojalvo and Anna Ritz and Greg Anderson and Miguel C. Soriano},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.07355},
eprint = {2609.07355},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.07355},
note = {CC BY-SA 4.0},
}
RIS
TY - UNPB
TI - Determinants of hyperparameter robustness in connectome reservoir computing
AU - Miles Walter Churchland
AU - Raul de Palma Aristides
AU - Jordi Garcia-Ojalvo
AU - Anna Ritz
AU - Greg Anderson
AU - Miguel C. Soriano
PY - 2026
DA - 2026/09/07
PB - arXiv
DO - 10.48550/arxiv.2609.07355
UR - https://arxiv.org/abs/2609.07355
AB - Reservoir computing provides a controlled setting for studying how recurrent network architectureshapes computation: input signals are projected into a high-dimensional state space by a fixed nonlinear dynamical system, and only the readout is trained. However, reservoir performance can be dependent on hyperparameters; this paper asks which recurrent network features support robustness to those parameter changes. We characterize computational performance using memory capacity (MC), truncated single-delay information-processing capacity (IPC), and kernel rank (KR). Generalization across input histories is measured using generalization rank (GR), while hyperparameter robustness is quantified using the coefficient of variation (CV) of each metric across sweeps of target spectral radius, input scaling, leak rate, and neuron bias. To examine the architectural determinants of robustness, we construct perturbations that alter connectivity topology, excitatory/inhibitory sign structure, weight magnitudes, and weight placement while preserving complementary properties. Across these experiments, the C. elegans connectome consistently occupies a relatively low-variance regime. The central result is a performance-robustness tradeoff: architecture variants with higher task-agnostic performance also tend to exhibit greater hyperparameter sensitivity and poorer common-tail generalization. Across the E/I edge balance sweeps and shuffle controls, this tradeoff is closely associated with the raw spectral radius before normalization. Because every perturbed matrix is rescaled to the same target radius, matrices with lower raw spectral radius receive greater global amplification of their recurrent weights. The observed differences among architectures therefore characterize the joint effects of structural variation and architecture-specific global rescaling under spectral-radius normalization.
ER -