LHCb Experiment Introduces Lamarr for Enhanced Simulation Efficiency
The LHCb experiment at CERN has introduced a new simulation framework named Lamarr, which utilizes machine learning models to enhance the speed and efficiency of data simulation. This development is crucial as approximately 90% of the computing resources for the LHCb experiment have been dedicated to generating simulated data samples for its second run. With the upcoming Run 3, the upgraded LHCb detector is expected to collect significantly larger data samples, necessitating an increase in the number of simulated events for effective data analysis.
Lamarr is built on the Gaudi framework and aims to streamline the simulation process by parameterizing both the detector response and the reconstruction algorithms. The framework employs Deep Generative Models that utilize various algorithms to effectively model the high-level responses of individual components of the LHCb detector. This includes encoding experimental errors and uncertainties that arise during detection and reconstruction phases into neural networks.
The models are trained using real data whenever feasible, which allows for the statistical subtraction of background components through appropriate reweighting procedures. By integrating Lamarr into the existing LHCb Gauss Simulation framework, the execution of simulations can be combined seamlessly with available generators, enhancing the overall simulation process.
The implications of this advancement are significant. As the demand for simulated data increases, Lamarr provides a solution that can potentially reduce the computational burden while maintaining the accuracy needed for high-energy physics experiments. This innovation not only supports the LHCb's objectives but also sets a precedent for future developments in simulation technologies within the field of experimental high-energy physics.
The findings and methodologies discussed in this paper are detailed in the work titled "Lamarr: LHCb ultra-fast simulation based on machine learning models deployed within Gauss" by Matteo Barbetti, available on arXiv under the identifier arXiv:2303.11428.