Kernaut's discovered kernel beats tuned standard kernels on glucose prediction
sirbayes · x · 2026-10-08
A follow-up on the Kernaut paper: using data from the UVA/Padova type-1 diabetes simulator for glucose-level prediction, the structured kernel discovered by Astra (the agent) outperforms the best tuned standard kernel.
Related event: Kernaut: Coding Agents Autodiscover Interpretable GP Kernels(3 posts)→
More from Research
- MIKE: researchers open-source compact post-quantum NIKE with 80-byte public keys — jedisct1 · 2026-10-08
- KITE transfers robot task knowledge across hands zero-shot, hitting 7/10 on real hardware — qinzytech · 2026-10-08
- NeurIPS paper: humanoids should imagine in 360° instead of thinking longer to find objects — _vztu · 2026-10-08
- Stanford's Anshul Kundaje Slams 'Universal Virtual Cell': Perturbation Prediction Is the Wrong Metric — anshulkundaje · 2026-10-08
- Kernaut: coding agents design Gaussian process kernels via program search — sirbayes · 2026-10-08
- Kernaut's discovered kernels stay interpretable: 16 scalar functions, inner-product form — sirbayes · 2026-10-08