Ganguli lab uses interpretable ML to identify emergent absence seizure mechanism
SuryaGanguli · x · 2026-09-20
Surya Ganguli's lab, led by Jacob Hull with John Huguenard, published a new paper using interpretable machine learning to identify an emergent absence seizure mechanism.
They built a highly accurate yet interpretable dynamical systems model of how the brain transitions into seizure and loses consciousness through distributed network chaos. The model predicts that when certain pairs of brain regions oscillate at a preferred phase lag, seizure amplitude grows — and not at the opposite lag. These predictions were confirmed with paired optogenetic stimulation across all phase lags, showing how neural recordings + ML models + interpretability can yield both conceptual insight and testable, nonintuitive predictions.
More from Research
- LLMs Can Detect If They're Batched via Floating-Point Summation Order — Sauers_ · 2026-10-25
- YC's next Paper Club on Nov 4 tackles AI for biology — ycombinator · 2026-10-25
- Optimal four-copy detector boosts three-tangle detection probability over 5x — MvsCerezo · 2026-10-25
- Chloroplast genomes proposed as the next frontier for generative DNA design — iskander · 2026-10-25
- Monadize fine-tune fixes Gemma's 'Elara problem': 60 unique names in 30 generations vs heavy repetition on base model — zoraruha · 2026-10-25
- KNOWS benchmark: best computer-use agent fully solves just 3.6% of tasks — anmarasovic · 2026-10-25