Speaker
Description
The unprecedented volume of data and Monte Carlo simulations at the HL-LHC poses increasing challenges for particle physics analyses, demanding computation-efficient analysis workflows and reduced time to insight. We present a review of data and statistical analysis models and tools in CMS, with a particular emphasis on the challenges and solutions associated with the recent W mass measurement. We present a comprehensive analysis framework that leverages RDataFrame, Eigen, Boost Histograms, and the Python scientific ecosystem, with particular emphasis on the interoperability between ROOT and Python tools and output formats (ROOT and HDF5). Our implementation spans from initial event processing to final statistical interpretation, featuring optimizations in C++ and RDataFrame that achieve favorable performance scaling for billions of events. The framework incorporates interfaces to TensorFlow for fast and accurate complex multi-dimensional binned maximum likelihood calculations and robust minimization. We will discuss gaps and deficiencies in standard ROOT-based tools and workflows, how these were addressed with alternative integrated or standalone solutions, and possible directions for future improvement.