We present CURTAINs, a fully data driven paradigm that improves on the weakly supervised searches. CURTAINs is designed to be sensitive to small density perturbations in n-dimensional feature space caused by the presence of signals. CURTAINs can be shown to be very robust in the absence of any signals, and yet be highly sensitive to signals even at very low signal to background...
A particularly interesting application of autoencoders (AE) for High Energy Physics is their use as anomaly detection (AD) algorithms to perform a signal-agnostic search for new physics. This is achieved by training AEs on standard model physics and tagging potential new physics events as anomalies. The use of an AE as an AD algorithm relies on the assumption that the network better...
In place of traditional cut and count methods, machine learning techniques offer powerful ways to optimise our searches for new physics. At the FCC-ee, we will probe the highest intensities and energies ever seen at a lepton collider, opening the possibly for discovery of massive right-handed neutrino states. In this work, existing searches for HNLs at the FCC-ee are optimised using a BDT and...
The ongoing Run-3 at the LHC is providing proton-proton collision data at the record energy of 13.6 TeV and, as of May 2024, a total integrated luminosity of almost 90 fb^-1 has been recorded by the ATLAS detector. This contribution presents preliminary results on the search for the pair production of stop squarks, the scalar supersymmetric partner of the top quark, based on the Run-2 and...
The Bert pretraining paradigm has proven to be highly effective in many domains including natural language processing, image processing and biology. To apply the Bert paradigm the data needs to be described as a set of tokens, and each token needs to be labelled. To date the Bert paradigm has not been explored in the context of HEP. The samples that form the data used in HEP can be described...
We present a novel approach for directly generating full events at detector-level from parton-level information, leveraging cutting-edge machine learning techniques. To address the challenge of multiplicity variations between parton and reconstructed object spaces, we employ transformers, score-based models and normalizing flows. Our method tackles the inherent complexities of the stochastic...
A dedicated experimental search for a muon electric dipole moment (EDM) is being set up at Paul Scherrer Institute. This experiment will search for a muon EDM signal with a final precision of \SI{6e-23}{e \cdot cm} using the frozen-spin technique, improving the current experimental limit by 3 orders of magnitude. To achieve the precision objective, it is important to optimize the setup to...
In this talk, we address the challenge of optimizing detector design for advanced tasks in high energy particle physics. Our goal is to develop differentiable pipelines for the optimization of typical metrics sought out in particle physics applications. The approach is tailored with a focus on several critical design aspects, including optimizing detector performance, enhancing sensitivity to...
The dynamic aperture is defined as the region of phase space where the particle motion in circular particle accelerators remains bounded over a fixed and large number of turns. Understanding the key features of this concept offers insight into non-linear beam dynamics and factors affecting beam lifetime in accelerators, which are pivotal for the operation of machines like the CERN Large Hadron...
Machine learning has become an indispensable tool in the field of high-energy physics, particularly in the CMS experiment at CERN.
In this talk, we will discuss some of the new developments in ML techniques implemented in the CMS experiment. These advancements have improved tasks like event reconstruction, jet flavour tagging, data quality monitoring, anomaly detection in the triggers, and...
In the ATLAS trigger and data acquisition system we can use machine learning to approximate existing online algorithms and accelerate trigger decisions in real time. This will be particularly important for the ATLAS Phase II upgrade in the high-luminosity LHC which will enforce strict latency requirements in the trigger. This work introduces a novel application of a Convolutional Neural...
Jets containing b-hadrons (b-jets) are a key signature to trigger events at collider experiments, as they're associated to many interesting physics processes, such as Higgs decays. Charged particle tracks reconstruction, the main input for b-tagging algorithms, makes the b-jet trigger selections some of the most CPU intensive ones at the ATLAS High-Level-Trigger (HLT). To cope with the...
Long lived particles (LLP) are ubiquitous in many Standard Model extensions, and could provide solutions to long-standing problems in modern physics. LLPs would present unique signatures, such as decays in flight far from the interaction point. New trigger and reconstruction techniques are required to search for such events. We propose using the LHCb muon detector as a sampling...
Leveraging the current industry shift towards heterogeneous computing and the widespread
adoption of FPGAs as accelerators to deploy machine learning algorithms, this project
introduces the Vitis accelerator backend, a novel backend for HLS4ML. HLS4ML is a python package tailored for machine learning inference on FGPAs that translates traditional
open-source machine learning package models...
The general-purpose Geant4 based calorimeter framework Lorenzetti Showers is being presented. It provides an ideal tool for simulating various configurations of calorimeter systems or testing their responses in complex scenarios. Its limits are being investigated by simulating a system very close to the ATLAS calorimeters and comparing their signatures under challenging conditions. Such a...
The Large-Sized Telescope (LST) is one of three telescope types being built as part of the Cherenkov Telescope Array Observatory (CTAO) to cover the lower energy range between 20 GeV and 200 GeV. The Large-Sized Telescope prototype (LST-1), installed at the La Palma Observatory Roque de Los Muchachos, is currently being commissioned and has successfully taken data since November 2019. The...
Imaging atmospheric Cherenkov telescopes (IACTs) observe extended air showers (EASs) initiated by the interaction of very-high-energy gamma rays and cosmic rays with the atmosphere. Cherenkov photons induced by a given shower are focused onto the camera plane of the telescopes resulting into a spatial and temporal charge development in the camera pixels. Besides the Cherenkov light emitted by...
Short-distance (SD) effects in $b\to s l^+l^-$ transitions can give large corrections to the Standard Model prediction. They can however not be computed from first principles. In my talk I will present a neural network, that takes such SD effects into account, when inferring the Wilson coefficients $C_9$ and $C_{10}$ from $b\to s l^+l^-$ angular observables. The model is based on...
The event-wise multi-dimensional unfolding is performed with the machine-learning-based OMNIFOLD algorithm to measure the event shape observables of the minimum bias data of low pile-up proton-proton collisions at a centre-of-mass energy of 13 TeV collected by the CMS detector. A machine-learning-based uncertainty estimation method is used to estimate the unbinned uncertainty and the...
Graph neural networks (GNNs) have recently emerged as state-of-the-art tools across various scientific disciplines due to their capability to represent complex relationships in datasets that lack simple spatial or sequential structures. This talk will explore the application of GNNs in two contrasting experimental environments. The first of which is the deep full event interpretation (dFEI) at...
The SND@LHC experiment aims to observe and measure neutrinos produced at the LHC. The goal of the detector reconstruction is therefore to identify events as coming from neutrinos against the typical large background from neutral hadrons and muons, and to identify the type of neutrino interaction. Current reconstruction methods are based on reconstructing muon tracks and rectangular cuts, and...
Data Quality Monitoring (DQM) is a crucial task in large particle physics experiments, since detector malfunctioning can compromise the data. DQM is currently performed by human shifters, which is costly and results in limited accuracy. In this work, we provide a proof-of-concept for applying human-in-the-loop Reinforcement Learning (RL) to automate the DQM process while adapting to operating...
This talk will summarize a method based on machine learning to play the devil's advocate and investigate the impact of unknown systematic effects in a quantitative way. This method proceeds by reversing the measurement process and using the physics results to interpret systematic effects under the Standard Model hypothesis. We explore this idea in arXiv:2303.15956 by considering the...
A good opportunity to ask the speakers all the things you were always afraid to ask.
