Zhiyuan Ning

SEP 3, 2026 · PREPRINT

Para-Pipe: Exploiting Hierarchical Operator Parallelism of ML Computational Graphs on SoCs

arXiv

This is a systems computer science paper presenting a novel optimization framework for edge ML inference, not a clinical or biomedical study; it reports engineering design and performance benchmarking on hardware without peer review.

SEP 3, 2026 · PREPRINT

RATL: Learning from Retrieved Residuals for Robust Multivariate Time-Series Forecasting

arXiv

This is a methodological computer science paper presenting a novel algorithmic approach (RATL) for time-series forecasting, without clinical or real-world validation data, comparative efficacy claims, or outcomes relevant to clinical practice.