Zebrafish brain circuits inspire energy-efficient and robust AI networks
Two microcircuits from zebrafish tectum boost AI efficiency and noise tolerance by 2026
A study of zebrafish tectal microcircuits identifies two functional subcircuits: ns_TIN, which shows a low spike footprint and measurable influence on prediction error, and superficial_TIN, which produces the highest robustness sensitivity. When transferred into ResNet18 for CIFAR-10, the ns_TIN-inspired module improves performance preservation under reduced computation, while the superficial_TIN-inspired module improves robustness under Gaussian noise. These findings provide a subcircuit-level route linking biological circuit organization with bio-inspired neural architecture design.
- Identified ns_TIN subcircuit as spike-efficient internal information processor (low spike footprint, high influence on prediction error).
- Identified superficial_TIN subcircuit as robustness-preserving feedback system (highest robustness sensitivity).
- Transferred motifs to ResNet18 on CIFAR-10: ns_TIN module preserved accuracy under reduced compute, superficial_TIN module improved noise resistance.
Why It Matters
Biological blueprints from zebrafish could lead to far more efficient and robust AI hardware and algorithms.