Description
Reception and research poster presentations
The Accelerated AI Algorithms for Data-Driven Discovery (A3D3) Institute, funded by the National Science Foundation (NSF), under the Harnessing the Data Revolution (HDR) program, is a multi-disciplinary and geographically distributed entity with the primary mission to lead a paradigm shift in the application of real-time artificial intelligence (AI) at scale to advance scientific knowledge and...
Given the recent advances of machine learning techniques, the Large Hadron Collider (LHC) at CERN is incorporating deep learning (DL) models, such as DeepCalo, to enhance the quality of data analysis of particle experiments. However, the need for in-time inference to keep up with data generation rates, as well as the dynamics of the experiments, require that the data processing feature short...
In this study, we demonstrate the potential of sparse point-voxel convolutional neural networks (SPVCNN) for hadronic calorimetry tasks using HCAL and HGCAL datasets. By employing a modified object condensation loss, we train the network to group cell deposits into clusters while filtering out noise. We show that SPVCNN performs comparably to generic topological cluster-based methods in both...
Efficient computational strategies are paramount for devices in resource-limited settings, particularly within high-energy physics experiments. To address this, we propose research primarily focused on improved energy efficiency and reduced latency inherent to AI algorithms implemented with analog circuits such as memristive crossbar arrays that perform in-memory matrix-vector multiply...
We present a denoising autoencoder for extracting low-energy signals in Liquid Argon Time Projection Chamber (LArTPC) detectors. In particular, we are interested in neutrinos originating from core-collapse supernova events, and the detection of these neutrinos can help improve our knowledge of the physics of core-collapse supernova events [1]. Additionally, if we can detect them fast enough,...
A graph neutral network (GNN) was constructed to identify charged lepton flavor violating decays of a tau particle into three muons in proton-proton collisions recorded with the CMS detector of the Large Hadron Collider. The muons from this decay are expected to have very low momentum, making them hard to detect in the high pileup environment expected at the high luminosity LHC (HL-LHC). We...
Recent studies on the ITk data showed that the Graph Neural Network (GNN) -based track finding can provide not only satisfied track efficiency but also reasonable track resolutions. However, the GNN-based track finding is computationally slow in CPUs, demanding the usage of coprocessors like GPUs to speed up the inference time. The large graph size, normally 300k nodes and 1M edges,...
Matched-filtering detection techniques for gravitational-wave (GW) signals in ground-based interferometers rely on having well-modeled templates of the GW emission.
However, interesting science cases aside from compact mergers do not yet have accurate enough modeling to make matched filtering possible, including core-collapse supernovae and other stochastic sources. Therefore the development...
At the LHC proton bunches are collided at a rate of 40MHz. The Compact Muon Superconducting Solenoid (CMS) detectorโs Level-1 (L1) trigger system is responsible for reducing this data rate to about 100kHz so that approximately 1% of these events can be saved for offline physics analyses. The task is to develop algorithms to determine what data to keep and what to discard. Traditionally,...
This poster presents an exploration into the realm of Beyond the Standard Model (BSM) Long-Lived Particles (LLPs) with a focus on the integration and development of a jet-tagging algorithm for the Compact Muon Solenoid (CMS) experiment's Level 1 (L1) Trigger system, in the context of the forthcoming upgrade to the High Luminosity Large Hadron Collider (HL-LHC). In spite of the challenges posed...
During the next update of the High-Luminosity Large Hadron Collider (HL-LHC) of ATLAS, a new global trigger subsystem will be installed into the L0 Trigger. New and improved hardware and algorithms will be deployed during the upgrade to increase the performance of the trigger system. The global trigger subsystem consists of various components, including the FPGA-based Global Event Processor...
Fast, accurate detector simulations are necessary to keep up with the data collected in the coming years in HEP. Due to their stochastic nature, ML-based generative models are natural opportunities for fast, differentiable simulations. We present two such graph- and attention-based models for generating LHC-like data using sparse and efficient point cloud representations, with state-of-the-art...
Machine learning on tiny IoT devices based on microcontroller units (MCU) is appealing but challenging: the memory of microcontrollers is 2-3 orders of magnitude smaller even than mobile phones. We propose MCUNet, a framework that jointly designs the efficient neural architecture (TinyNAS) and the lightweight inference engine (TinyEngine), enabling ImageNet-scale inference on microcontrollers....
On-device training enables the model to adapt to new data collected from the sensors by fine-tuning a pre-trained model. However, the training memory consumption is prohibitive for IoT devices that have tiny memory resources. We propose an algorithm-system co-design framework to make on-device training possible with only 256KB of memory. On-device training faces two unique challenges: (1) the...
The detection of a supernova burst is a unique opportunity to derive insights on astro and particle physics especially neutrinos. Neutrinos are the first hint of a supernova occuring to arrive on Earth due to their very low interaction cross section. They can provide extremely valuable information on the direction of burst enabling to point optical detection systems there in a multi messenger...
The particle-flow (PF) algorithm, which infers particles based on tracks and calorimeter clusters, is of central importance to event reconstruction in the CMS experiment at the CERN LHC, and has been a focus of development in light of planned Phase-2 running conditions with an increased pileup and detector granularity. In recent years, the machine learned particle-flow (MLPF) algorithm, a...
The Particle Flow algorithm has proven highly effective in the offline reconstruction of events in the CMS detector. Combined with Pile-Up Per Particle Identification (PUPPI), the two algorithms provide the necessary basis for the construction of higher-level physics options, such as jets and taus. With the upcoming High Luminosity upgrade of the Large Hadron Collider (HL-LHC), implementing...
Increased development and utilization of multimodal scanning probe microscopy (SPM) and spectroscopy techniques have led to an orders-of-magnitude increase in the volume, velocity, and variety of collected data. While larger datasets have certain advantages, practical challenges arise from their increased complexity including the extraction and analysis of actionable scientific information. In...
Limited by the lack of truth labels on real data, fully supervised ML algorithms are constrained to training only with simulated samples. With self-supervised learning, we can leverage vast amounts of unlabeled real data to facilitate training. We investigate the application of [VICReg][1], a contrastive learning model, on a classification task: discriminating signal jets (e.g. $H \rightarrow...
Spectrograms are frequently used to provide qualitative insights into the types of noise and signals present in audio data. Similarly, we can use them to gain insights from data such as real gravitational wave from gravitational wave detectors. Simply by eye, we can see characteristic chirp signals from gravitational waves due to the physics of the black holes' inspiral. Designing a novel...
The Laser Interferometer Gravitational Wave-Observatory (LIGO) has accumulated more than 4.5 petabytes (Pb) of data in its quest to detect gravitational waves. Furthermore, it is anticipated that the total data accrued will increase by approximately 0.8 petabytes per year. The processing and analysis of the extensive volume of data from LIGO necessitates a tremendous amount of computational...
Brain-computer interfaces use the electrical activity of the brain to control an external device, but decoding complex neural signals requires large amounts of computational power and time. We use a novel convex optimization algorithm to do real-time feature selection based on relevance, sparsity, and smoothness. We demonstrate that the algorithm can reduce the feature set while maintaining...
The Compact Muon Solenoid (CMS) detector is one of two general-purpose detectors at
the CERN LHC. Products of proton-proton collisions at a center of mass energy of 13 TeV are reconstructed in the CMS detector to probe the standard model of particle physics and to search for processes beyond the standard model. The development
of precision algorithms for this reconstruction is therefore a...
