arXiv subjects

Valentina Guglielmi

Publications and source records attributed to Valentina Guglielmi.

3 recordsLinked to original sources

arXiv2026

Knowledge Distillation of a Normalising Flow for Real-Time Anomaly Detection at LHC Level-1 Trigger

We present a new strategy for unsupervised anomaly detection at the hardware-based first stage of event processing (Level-1 trigger) of the Large Hadron Collider (LHC). A normalising flow trained exclusively on Standard Model (SM) events provides a teacher anomaly score based on its exact negative log-likelihood, which is distilled into a compact neural network suitable for field-programmable gate array (FPGA) deployment. The teacher reaches state-of-the-art performance on a benchmark dataset, using a rigorous p-value-based definition of the anomaly score. A simple two-hidden-layer student reproduces the teacher performance across four beyond-the-SM benchmarks to within 0.1 percentage points in area under the receiver operating characteristic curve (AUC), while achieving a compression factor of approximately 325 times. Quantisation-aware training with PQuantML reduces the model to 8-bit precision with AUC changes below 0.06 percentage points. Compilation to Verilog firmware with Alkaid yields FPGA designs that require neither digital signal processors nor block random-access memory and achieve latencies of 27-52 ns. This modular pipeline decouples the expressiveness of the anomaly-detection model from real-time hardware constraints, enabling complex architectures to be used for model-agnostic searches for new physics at the full LHC collision rate of 40 MHz.

hep-ex↗

arXiv2023

Machine learning approaches for parameter reweighting in MC samples of top quark production in CMS

In particle physics, Monte Carlo (MC) event generators are needed to compare theory to the measured data. Many MC samples have to be generated to account for theoretical systematic uncertainties, at a significant computational cost. Therefore, the MC statistic becomes a limiting factor for most measurements and the significant computational cost of these programs a bottleneck in most physics analyses. In this contribution, the Deep neural network using Classification for Tuning and Reweighting (DCTR) approach is evaluated for the reweighting of two systematic uncertainties in MC simulations of top quark pair production within the CMS experiment. DCTR is a method, based on a Deep Neural Network (DNN) technique, to reweight simulations to different model parameters by using the full kinematic information in the event. This methodology avoids the need for simulating the detector response multiple times by incorporating the relevant variations in a single sample.

hep-ex↗

arXiv2022

Machine learning approaches for parameter reweighting in Monte-Carlo samples of top quark production in CMS

In high-energy particle physics, complex Monte Carlo (MC) simulations are needed to compare theory predictions to measurable quantities. Many and large MC samples are needed to be generated to take into account all the systematics. Therefore, the MC statistics (and hence the MC modeling uncertainties) become a limiting factor for most measurements. Moreover, the significant computational cost of these programs becomes a bottleneck in most physics analyses. Therefore, it is extremely important to find a way to reduce the MC samples generated to decrease the MC statistical uncertainties and lower the computational cost. In these proceedings, we evaluate an approach called Deep neural network using Classification for Tuning and Reweighting (DCTR). DCTR is a method based on a Deep Neural Network (DNN) to reweight simulations to different models or model parameters and fit simulations, using the full kinematic information in the event. This reweighting methodology avoids the need for simulating the detector response multiple times by incorporating the relevant variations in a single sample. In this way, the MC statistical uncertainties and the computational cost are both reduced. Moreover, unlike the standard reweighting, in which the ratio in bins of two histograms at truth level is performed, multidimensional and unbinned information can be used as inputs to the DNN. In addition, DCTR can perform tasks that are not possible with other current existing methods, such as continuous reweighting as a function of any MC parameter, simultaneous reweighting of more MC parameters and tuning MC simulations to the data. We test the method on MC simulations of top quark pair production, which we reweight to different SM parameter values and to different QCD models.

hep-ex↗