Event Classification with Multi-step Machine Learning

The usefulness and value of Multi-step Machine Learning (ML), where a task is organized into connected sub-tasks with known intermediate inference goals, as opposed to single large model learned end-to-end without sub-tasks, presented. Pre-optimized ML models are better performance obtained by re-optimizing the one. selection an from several small candidates for each sub-task has been performed using idea based on Neural Architecture Search (NAS). In this paper, Differentiable (DARTS) Single Path One-Shot NAS (SPOS-NAS) tested, construction loss functions improved keep all smoothly learning. Using DARTS SPOS-NAS optimization well connections multi-step machine learning systems, we find that (1) such system can quickly successfully select highly performant combinations, (2) selected consistent baseline algorithms, grid search, their outputs controlled.