A5039198154- Particle physics theoretical and experimental studies
- Particle Detector Development and Performance
- Computational Physics and Python Applications
- Distributed and Parallel Computing Systems
- Scientific Computing and Data Management
- Pulsars and Gravitational Waves Research
- Parallel Computing and Optimization Techniques
- High-Energy Particle Collisions Research
- Gamma-ray bursts and supernovae
- Dark Matter and Cosmic Phenomena
- Neural Networks and Applications
- Advanced Neural Network Applications
- Advanced Data Storage Technologies
- Anomaly Detection Techniques and Applications
- Research Data Management Practices
- Seismology and Earthquake Studies
- Machine Learning in Materials Science
- Image and Signal Denoising Methods
- Quantum Chromodynamics and Particle Interactions
- Atomic and Subatomic Physics Research
- Machine Learning and Algorithms
- Blind Source Separation Techniques
- CCD and CMOS Imaging Sensors
- Model Reduction and Neural Networks
- Machine Learning and Data Classification
Massachusetts Institute of Technology
2019-2024
Moscow Institute of Thermal Technology
2023-2024
IIT@MIT
2022-2024
The NSF AI Institute for Artificial Intelligence and Fundamental Interactions
2021
Universität Hamburg
2019
Liverpool John Moores University
1996
The Open University
1996
Jet substructure has emerged to play a central role at the Large Hadron Collider, where it provided numerous innovative ways search for new physics and probe Standard Model, particularly in extreme regions of phase space. In this article we focus on review development use state-of-the-art jet techniques by ATLAS CMS experiments.
Abstract Compact symbolic expressions have been shown to be more efficient than neural network (NN) models in terms of resource consumption and inference speed when implemented on custom hardware such as field-programmable gate arrays (FPGAs), while maintaining comparable accuracy (Tsoi et al 2024 EPJ Web Conf. 295 09036). These capabilities are highly valuable environments with stringent computational constraints, high-energy physics experiments at the CERN Large Hadron Collider. However,...
We introduce an automated tool for deploying ultra low-latency, low-power deep neural networks with convolutional layers on FPGAs. By extending the hls4ml library, we demonstrate inference latency of $5\,\mu$s using architectures, targeting microsecond applications like those at CERN Large Hadron Collider. Considering benchmark models trained Street View House Numbers Dataset, various methods model compression in order to fit computational constraints a typical FPGA device used trigger and...
A foundational set of findable, accessible, interoperable, and reusable (FAIR) principles were proposed in 2016 as prerequisites for proper data management stewardship, with the goal enabling reusability scholarly data. The also meant to apply other digital assets, at a high level, over time, FAIR guiding have been re-interpreted or extended include software, tools, algorithms, workflows that produce are now being adapted context AI models datasets. Here, we present perspectives, vision,...
Abstract Construction of an e + - Higgs factory has been identified as a major goal for particle physics. Such collider will offer precise measurements the bosons couplings to other particles. A extendable in energy can also establish self-coupling, measure coupling top quark, and expand reach probe new phenomena. We propose strategy energy-extendable based on linear accelerator technology. This offers compact cost-effective design that could be realized project US. The core technologies...
Machine Learning is a powerful tool to reveal and exploit correlations in multi-dimensional parameter space. Making predictions from such highly non-trivial task, particular when the details of underlying dynamics theoretical model are not fully understood. Using adversarial networks, we include priori known sources systematic uncertainties during training. This paves way more reliable event classification on an event-by-event basis, as well novel approaches perform fits particle physics...
We present the implementation of binary and ternary neural networks in hls4ml library, designed to automatically convert deep network models digital circuits with FPGA firmware. Starting from benchmark trained floating point precision, we investigate different strategies reduce network's resource consumption by reducing numerical precision parameters or ternary. discuss trade-off between model accuracy consumption. In addition, show how balance latency retaining full on a selected subset...
A bstract Discoveries of new phenomena often involve a dedicated search for hypothetical physics signature. Recently, novel deep learning techniques have emerged anomaly detection in the absence signal prior. However, by ignoring priors, sensitivity these approaches is significantly reduced. We present strategy dubbed Quasi Anomalous Knowledge (QUAK), whereby we introduce alternative priors that capture some salient features signatures, allowing recovery even when incorrect. This approach...
Despite advances in the programmable logic capabilities of modern trigger systems, a significant bottleneck remains amount data to be transported from detector off-detector where decisions are made. We demonstrate that neural network autoencoder model can implemented radiation tolerant ASIC perform lossy compression alleviating transmission problem while preserving critical information energy profile. For our application, we consider high-granularity calorimeter CMS experiment at CERN Large...
Abstract Matched-filtering detection techniques for gravitational-wave (GW) signals in ground-based interferometers rely on having well-modeled templates of the GW emission. Such have been traditionally used searches compact binary coalescences (CBCs), and employed all known detections so far. However, interesting science cases aside from mergers do not yet accurate enough modeling to make matched filtering possible, including core-collapse supernovae sources where stochasticity may be...
To enable the reusability of massive scientific datasets by humans and machines, researchers aim to adhere principles findability, accessibility, interoperability, (FAIR) for data artificial intelligence (AI) models. This article provides a domain-agnostic, step-by-step assessment guide evaluate whether or not given dataset meets these principles. We demonstrate how use this FAIRness an open simulated produced CMS Collaboration at CERN Large Hadron Collider. consists Higgs boson decays quark...
Abstract Low-latency noise regression algorithms are crucial for maximizing the science outcomes of LIGO, Virgo, and KAGRA gravitational-wave detectors. This includes improvements in detectability, source localization pre-merger detectability signals thereby enabling rapid multi-messenger follow-up. In this paper, we demonstrate effectiveness DeepClean , a convolutional neural network architecture that uses witness sensors to estimate subtract non-linear non-stationary from strain data. Our...
Technical and environmental noise in ground-based laser interferometers designed for gravitational-wave observations like Advanced LIGO, Virgo KAGRA, can manifest as narrow (<1Hz) or broadband ($10'$s even $100'$s of Hz) spectral lines features the instruments' strain amplitude density. When sources this cannot be identified removed, cases where there are witness sensors sensitive to source, denoising channel performed software, enabling recovery instrument sensitivity over affected...
We develop and study FPGA implementations of algorithms for charged particle tracking based on graph neural networks. The two complementary designs are OpenCL, a framework writing programs that execute across heterogeneous platforms, hls4ml, high-level-synthesis-based compiler network to firmware conversion. evaluate compare the resource usage, latency, performance our benchmark dataset. find considerable speedup over CPU-based execution is possible, potentially enabling such be used...
Abstract In the next decade, demands for computing in large scientific experiments are expected to grow tremendously. During same time period, CPU performance increases will be limited. At CERN Large Hadron Collider (LHC), these two issues confront one another as collider is upgraded high luminosity running. Alternative processors such graphics processing units (GPUs) can resolve this confrontation provided that algorithms sufficiently accelerated. many cases, algorithmic speedups found...
In this paper, we present a method of embedding physics data manifolds with metric structure into lower dimensional spaces simpler metrics, such as Euclidean and Hyperbolic spaces. We then demonstrate that it can be powerful step in the analysis pipeline for many applications. Using progressively more realistic simulated collisions at Large Hadron Collider, show approach learns underlying latent structure. With notion volume spaces, provide first time viable solution to quantifying true...
Abstract In this paper, we investigate how field programmable gate arrays can serve as hardware accelerators for real-time semantic segmentation tasks relevant autonomous driving. Considering compressed versions of the ENet convolutional neural network architecture, demonstrate a fully-on-chip deployment with latency 4.9 ms per image, using less than 30% available resources on Xilinx ZCU102 evaluation board. The is reduced to 3 image when increasing batch size ten, corresponding use case...
Abstract Recurrent neural networks have been shown to be effective architectures for many tasks in high energy physics, and thus widely adopted. Their use low-latency environments has, however, limited as a result of the difficulties implementing recurrent on field-programmable gate arrays (FPGAs). In this paper we present an implementation two types network layers—long short-term memory gated unit—within hls4ml framework. We demonstrate that our is capable producing designs both small large...
The promise of multi-messenger astronomy relies on the rapid detection gravitational waves at very low latencies ($\mathcal{O}$(1\,s)) in order to maximize amount time available for follow-up observations. In recent years, neural-networks have demonstrated robust non-linear modeling capabilities and millisecond-scale inference a comparatively small computational footprint, making them an attractive family algorithms this context. However, integration these into gravitational-wave...