A5057362506- Particle physics theoretical and experimental studies
- Cosmology and Gravitation Theories
- Black Holes and Theoretical Physics
- Computational Physics and Python Applications
- High-Energy Particle Collisions Research
- Particle Detector Development and Performance
- Anomaly Detection Techniques and Applications
- Quantum Chromodynamics and Particle Interactions
- Gaussian Processes and Bayesian Inference
- Astrophysics and Cosmic Phenomena
- Algorithms and Data Compression
- Bayesian Methods and Mixture Models
- Medical Imaging Techniques and Applications
- Particle Accelerators and Free-Electron Lasers
- Superconducting and THz Device Technology
- Cold Atom Physics and Bose-Einstein Condensates
- Distributed and Parallel Computing Systems
- Radio Astronomy Observations and Technology
- Superconducting Materials and Applications
- Dark Matter and Cosmic Phenomena
- Generative Adversarial Networks and Image Synthesis
- COVID-19 diagnosis using AI
- Atomic and Subatomic Physics Research
- Astrophysical Phenomena and Observations
- International Science and Diplomacy
University of Ulster
2024
Heidelberg University
2021-2023
Jožef Stefan Institute
2019-2022
Consejo Nacional de Investigaciones Científicas y Técnicas
2022
National University of General San Martín
2022
International Center for Advanced Studies
2022
Heidelberg Institute for Theoretical Studies
2021
University of Plymouth
2017-2019
University of Sussex
2015-2017
Based on the established task of identifying boosted, hadronically decaying top quarks, we compare a wide range modern machine learning approaches. Unlike most methods they rely low-level input, for instance calorimeter output. While their network architectures are vastly different, performance is comparatively similar. In general, find that these new approaches extremely powerful and great fun.
We apply techniques from Bayesian generative statistical modeling to uncover hidden features in jet substructure observables that discriminate between different a priori unknown underlying short distance physical processes multijet events. In particular, we use mixed membership model known as latent Dirichlet allocation build data-driven unsupervised top-quark tagger and $t\overline{t}$ event classifier. compare our proposal existing traditional machine learning approaches top-jet tagging....
Given the disappearance of 750 GeV di-photon LHC signal and absence signals at high mass in this other channels, significant constraints on mixed Higgs-radion five-dimensional Randall-Sundrum model arise. By combining all these place a radion-mass-dependent lower bound radion vacuum expectation value that is fairly independent amount Higgs mixing.
A bstract We describe a technique to learn the underlying structure of collider events directly from data, without having particular theoretical model in mind. It allows infer aspects that may have given rise this structure, and can be used cluster or classify for analysis purposes. The unsupervised machine-learning is based on probabilistic (Bayesian) generative Latent Dirichlet Allocation. pair with an approximate inference algorithm called Variational Inference, which we then use extract...
Autoencoders as tools behind anomaly searches at the LHC have structural problem that they only work in one direction, extracting jets with higher complexity but not other way around. To address this, we derive classifiers from latent space of (variational) autoencoders, specifically Gaussian mixture and Dirichlet spaces. In particular, setup solves improves both performance interpretability networks.
Unsupervised anomaly-detection could be crucial in future analyses searching for rare phenomena large datasets, as example collected at the LHC. To this end, we introduce a physics inspired variational autoencoder (VAE) architecture which performs competitively and robustly on LHC Olympics Machine Learning Challenge datasets. We demonstrate how embedding some physical observables directly into VAE latent space, while same time keeping manifestly agnostic to them, can help identify...
Searches for anomalies are a significant motivation the LHC and help define key analysis steps, including triggers. We discuss specific examples how can be defined through probability density estimates, evaluated in physics space or an appropriate neural network latent space, model-dependence choosing data parameterisation. illustrate this classical k-means clustering, Dirichlet variational autoencoder, invertible networks. For two especially challenging scenarios of jets from dark sector we...
Collider searches face the challenge of defining a representation high-dimensional data such that physical symmetries are manifest, discriminating features retained, and choice is new-physics agnostic. We introduce JetCLR to solve mapping from low-level optimized observables through self-supervised contrastive learning. As an example, we construct for top QCD jets using permutation-invariant transformer-encoder network visualize its symmetry properties. compare with alternative...
Autoencoders are an effective analysis tool for the LHC, as they represent one of its main goal finding physics beyond Standard Model. The key challenge is that out-of-distribution anomaly searches based on compressibility features do not apply to while existing density-based lack performance. We present first autoencoder which identifies anomalous jets symmetrically in directions higher and lower complexity. normalized combines a standard bottleneck architecture with well-defined...
We develop a self-supervised method for density-based anomaly detection using contrastive learning, and test it event-level data from CMS ADC2021. The AnomalyCLR technique is data-driven uses augmentations of the background to mimic non-Standard-Model events in model-agnostic way. It permutation-invariant Transformer Encoder architecture map objects measured collider event representation space, where define space which sensitive potential anomalous features. An AutoEncoder trained on...
We report on the status of efforts to improve reinterpretation searches and measurements at LHC in terms models for new physics, context Reinterpretation Forum. detail current experimental offerings direct particles, measurements, technical implementations Open Data, provide a set recommendations further improving presentation results order better enable future. also brief description existing software frameworks recent global analyses physics that make use data.
We investigate a method of model-agnostic anomaly detection through studying jets, collimated sprays particles produced in high-energy collisions. train transformer neural network to encode simulated QCD "event space" dijets into low-dimensional "latent representation. optimize the using self-supervised contrastive loss, which encourages preservation known physical symmetries dijets. then binary classifier discriminate BSM resonant dijet signal from background both event space and latent...
We study the finite-temperature properties of Randall-Sundrum model in presence brane-localized curvature. At high temperature, as dictated by AdS/CFT, theory is a confined phase dual to planar AdS black hole. When radion stabilized, \`a la Goldberger-Wise, holographic first-order transition proceeds. The curvature contributes kinetic energy, substantially decreasing critical bubble energy. Contrary previous results, completes at much larger values $N$, number degrees freedom CFT. Moreover,...
The clockwork mechanism has recently been proposed as a natural way to generate hierarchies among parameters in quantum field theories. is characterized by very specific pattern of spontaneous and explicit symmetry breaking, the presence new light states referred `gears'. In this paper we begin investigating self-interactions these gears scalar model find parity-like selection rule at all orders fields. We then proceed investigate how can be realized 5D linear dilaton models from breaking...
A bstract We introduce a set of clockwork models flavor that can naturally explain the large hierarchies Standard Model quark masses and mixing angles. Since only contains chains new vector-like fermions without any other dynamical fields, constraints allow for relatively light physics scale. For two benchmarks with gear just above 1 TeV, allowed by constraints, we discuss collider searches possible ways reconstructing spectra at LHC. also examine similarities differences common solutions to...
Discriminating quark-like from gluon-like jets is, in many ways, a key challenge for LHC analyses. First, we use known difference PYTHIA and HERWIG simulations to show how decorrelated taggers would break down when the most distinctive feature is aligned with theory uncertainties. We propose conditional training on interpolated samples, combined controlled Bayesian network, as more resilient framework. The interpolation parameter can be used optimize evaluated calibration dataset, test...
In this work we study constraints from new searches for heavy particles at the LHC on allowed masses and couplings of a Kaluza-Klein (KK) graviton in holographic composite Higgs model. Keeping electroweak states such that precision tests are satisfied, control mass lightest KK using brane kinetic term. With 0.5--3 TeV. our analysis also employ little Randall-Sundrum (RS) models, characterized by lower UV scale five-dimensional model which turn implies modified to massless bulk fields....
Optimisation problems are ubiquitous in particle and astrophysics, involve locating the optimum of a complicated function many parameters that may be computationally expensive to evaluate. We describe number global optimisation algorithms not yet widely used benchmark them against random sampling existing techniques, perform detailed comparison their performance on range test functions. These include four analytic functions varying dimensionality, realistic example derived from recent fit...
Even though jet substructure was not an original design consideration for the Large Hadron Collider (LHC) experiments, it has emerged as essential tool current physics program. We examine role of on motivation and future energy Frontier colliders. In particular, we discuss need a vibrant theory experimental research development program to extend into new regimes probed by Jet organically evolved with close connection between theorists experimentalists catalyzed exciting innovations in both...
Simple Composite Higgs models predict new vector-like fermions not too far from the electroweak scale, yet LHC limits are now sensitive to TeV scale. Motivated by this tension, we explore holographic dual of minimal model, MCHM5, try and alleviate tension without increasing fine-tuning in potential. Interestingly, find that lowering UV cutoff 5D picture allows for heavier top partners less fine-tuning. In 4D corresponds number "colours" N , thus decay constant Goldstone Higgs. This is...
Collider searches face the challenge of defining a representation high-dimensional data such that physical symmetries are manifest, discriminating features retained, and choice is new-physics agnostic. We introduce JetCLR to solve mapping from low-level optimized observables though self-supervised contrastive learning. As an example, we construct for top QCD jets using permutation-invariant transformer-encoder network visualize its symmetry properties. compare with alternative...