Published on 15 May 2024
Interactions and complex dynamics of microbubble contrast agents
This paper studies a mathematical model of three interacting microbubble contrast agents, used in biomedical ultrasound imaging. It demonstrates a variety of complex dynamics like chaos and hyperchaos emerging from the interactions of the bubbles. Key control parameters tuning the dynamics are identified.
Published on 15 May 2024
AGN jet and star formation connections in nearby galaxies
This paper introduces a project investigating connections between active galactic nucleus (AGN) jet activity, gas outflows, and star formation rates (SFRs) in nearby galaxies. The goal is to determine if powerful jets suppress or enhance star formation. The sample consists of 35 nearby AGNs with and without powerful jets. Submillimeter observations from SCUBA-2 provide robust SFRs from spectral energy distribution fitting. Spatially-resolved optical/near-infrared spectroscopy measures outflow properties. Initial results show SFRs are overest...
Published on 15 May 2024
Human and AI teamwork improves knowledge graph analysis
This paper proposes a human-AI team (HAIT) system called KG-HAIT to improve knowledge graph analysis, specifically link prediction, through collaboration. Humans design an insightful feature extraction methodology leveraging dynamic programming on the graph. By integrating these human-derived features into AI training, KG-HAIT achieved significant performance gains across benchmarks, showcasing the power of human conceptual analysis combined with AI scalability.
Published on 15 May 2024
Probing the size and motion of a quasar's emission region using gravitational lensing
This paper analyzes distortions in the carbon IV emission line profile in two gravitationally lensed quasars to probe the geometry, size and motion of the broad line emission region around the quasars' supermassive black holes. Simple models of a rotating disk, wind, or jet emission region are compared to observations. The disk models best fit the data, with estimated emission region sizes of 5-7 and 5-12 light days across. These sizes fit with previous luminosity-radius relations. The sizes are systematically smaller than previous technique...
Published on 15 May 2024
Using machine learning for gravitational wave detection despite noise artifacts
This paper leverages the inherent robustness of machine learning models to reliably infer gravitational wave source parameters even when unknown 'glitches' (noise transients) contaminate the signal. Their normalizing flow network provides rapid parameter estimates without needing to first model and remove glitches. They investigate factors impacting result accuracy, finding the glitch properties themselves primarily affect robustness. With improvements, this method could accelerate electromagnetic counterpart search and provide deglitching p...
Published on 15 May 2024
Model for electron turbulence transport
Researchers derived a predictive model for electron heat transport driven by electron temperature gradient turbulence in tokamak fusion devices. Their model is a scaling law that depends on magnetic geometry and screening effects in addition to temperature/density gradients. They tested predictions extensively, even beyond training data, quantifying uncertainty. Overall it competes with or outperforms existing models.
Published on 15 May 2024
Dimensional reduction of Chern-Simons gauge theories yields quantum mechanical models
This paper studies dimensional reduction of 3d Chern-Simons gauge theories with spherical symmetry. For pure Chern-Simons, the reduced 1d theory resembles a fermionic model, and a duality relates the 3d and 1d quantum theories. For Chern-Simons-Higgs, the reduced model is a supersymmetric quantum mechanics governing the monopoles.
Published on 15 May 2024
Assessing Super-Resolution Image Quality
This paper proposes a dual-branch reduced-reference network called PFIQA to evaluate the perceptual quality and reconstruction fidelity of super-resolution images, without needing high-resolution reference images. It leverages vision transformers and CNNs to extract comprehensive visual features, incorporates scale factors to align with human perception, and uses patch-level scoring to enable fine-grained assessment.
Published on 15 May 2024
Charged scalar particle dynamics in a magnetic universe
This paper explores the behavior of charged scalar particles in an unusual gravitational environment containing a strong magnetic field aligned along the axis of symmetry. Using advanced quantum mechanics formalism, the authors analyze effects on particle motion, scattering, and quantized energy levels. Key findings relate to bound state solutions, accumulation of high energy states, and symmetry between particles and anti-particles.
Published on 15 May 2024
Attacking speech recognition via style transfer
This paper proposes a new method to attack automatic speech recognition systems by leveraging style transfer techniques. It allows attackers to generate adversarial examples that are customized to user-selected styles, making them harder to detect. The attack is done in two ways: directly adding perturbations to stylized audio, or optimizing the internal style codes. Experiments show over 80% success rate in targeted attacks, while maintaining naturalness.
Published on 15 May 2024
Zebrafish collective behavior changes studied in 3D
Researchers studied groups of 50 zebrafish swimming in a 3D environment, tracking their positions over time. As the fish adapted to new conditions, their collective behavior changed between more ordered and more random. Two key length scales captured these changes: the persistence length of motion before reorienting, and the nearest neighbor distance between fish. The ratio of these lengths correlated with group polarization explained by a model of self-propulsion with alignment interactions.
Published on 15 May 2024
Predicting neutron star merger gravitational wave spectra
This paper investigates using artificial neural networks to predict gravitational wave spectra in the post-merger phase of neutron star mergers. These predictions depend on neutron star mass, tidal deformability, and radius gradient. The neural network performed better than a baseline multiple linear regression method in cross-validation. Recalibrating the predictions based on peak frequency alignment shows the potential for more accurate predictions when empirical relations improve.
Published on 15 May 2024
Possible binding of nucleon, D meson, and D* meson
This paper investigates whether a three-body system composed of a nucleon, D meson, and D* meson can form a bound state. The authors use a framework called the Fixed Center Approximation, which models the system as a cluster of two particles (here either the ND* or ND pair) with the third particle (D or D* respectively) interacting with that cluster. They find there is enough attraction to bind the DND* system by about 60 MeV below the Lambda_c* D threshold, with a width around 90 MeV. If observed, these predicted states could provide valuab...
Published on 15 May 2024
Measuring entanglement in three-qubit systems
Researchers introduced new geometric methods to quantify global and genuine entanglement in three-qubit quantum systems. By mapping quantum states to an 'entanglement polytope' geometric space, they defined measures based on projecting and rejecting points onto biseparable line segments. They also showed how solving the 'inverse problem' allows controlling entanglement by finding states that match desired geometric properties.
Published on 15 May 2024
Auctioning future blocks improves builder rewards
This paper proposes an auction mechanism called Flashback where blockchain builders bid high-value transaction bundles to future block proposers in advance to improve profits. Simulations show Flashback builders can earn 20% higher rewards than regular builders.
Published on 15 May 2024
Scheduling policies for global quantum satellite networks
This paper investigates transmission scheduling algorithms to coordinate connections between satellites and ground stations for distributing quantum entanglement. The authors prove the general scheduling problem is computationally intractable, so they propose and assess four heuristic algorithms using simulations of the Starlink megaconstellation. They find algorithms based on optimal weighted matching and global greedy heuristics perform best across metrics like distribution rate, fidelity, and handover cost as ground station density increa...
Published on 15 May 2024
Using expert gaze to improve detection of vulvovaginal candidiasis
This paper introduces Gaze-DETR, a new method to improve the accuracy of neural networks in detecting vulvovaginal candidiasis (VVC) in images. VVC has sparse distribution and visually ambiguous features that make identification difficult. The key insight is that areas focused on yet not marked by experts are aligned with false positives from models. Gaze-DETR uses expert gaze data to enhance precision by reducing these false positives. It has a gaze-guided warmup to handle imbalanced VVC quantities and a rectification strategy to focus on c...
Published on 15 May 2024
Suppressing gauge drift with projections
This paper proposes a method to maintain gauge invariance in quantum simulations of lattice gauge theories, which is challenging due to inevitable errors producing unphysical states over time. It uses frequent projection of states back into the physical subspace, leveraging the quantum Zeno effect, along with random gauge transformations to reduce drift speed and projection frequency. The combined technique is demonstrated successfully on a simple quantum simulation.
Published on 15 May 2024
Real space and momentum space views reveal 3D nature of charge density waves in quasi-1D CuTe
Researchers present multiple techniques to fully characterize charge density waves (CDWs) in the quasi-one-dimensional material CuTe. Direct imaging shows periodic lattice distortions consistent with 3D CDW ordering. Angle-resolved photoemission spectroscopy reveals out-of-plane momentum dependence of the CDW gap size, confirming 3D electronic modulations. Time-resolved measurements identify coherent oscillations corresponding to theoretically predicted soft phonon modes driving the CDW formation.
Published on 15 May 2024
Dissipative trapping in noisy harmonic potentials
This paper introduces a model called the OU^2 process to study particles diffusing in harmonic potentials with stochastic stiffness fluctuations, as arises in optical tweezers experiments. Analytical and numerical methods reveal altered trapping behavior and statistics compared to the standard Ornstein-Uhlenbeck process, including power law tails in position distribution. Key results relate to entropy production, extreme value statistics, first passage times, and homogenization theories.
Published on 15 May 2024
Glass boundary detection for segmentation
This paper proposes a new deep learning approach to glass segmentation that focuses on detecting glass boundaries and avoiding over-capturing spurious features. A wide, shallow network architecture is used to extract large-scale glass regions while also embedding strong boundary constraints. Attention mechanisms filter noise and supplement detail within segmented regions. Experiments show state-of-the-art performance on multiple datasets.
Published on 15 May 2024
Microwave sensing for lung disease detection
This paper presents a new method to detect five common lung diseases - asthma, COPD, interstitial lung disease, pneumonia, and tuberculosis - using software-defined radio signals at 5.23 GHz reflected off patients' chests. The distinct breathing patterns caused by each disease modulate the signals differently. Data was collected from 116 patients and healthy subjects. A convolutional neural network model achieved 97% accuracy in classifying diseases, with high precision and recall scores. The system only needs data from 7 frequencies for 96%...
Published on 15 May 2024
Counting polymer coverings on rectangular lattices
This paper investigates the problem of counting configurations of rigid linear polymers covering adjacent sites on 2D rectangular lattices, leaving some sites unoccupied (monomers). It proves these coverings satisfy general recurrence relations for arbitrary polymer length and lattice width. This includes the #P-complete monomer-dimer problem. The recurrences could provide insights into solving long-standing complexity problems.
Published on 15 May 2024
Photonic Landau levels created in a laser cavity
This paper proposes a new method to create and control photonic Landau levels, which are quantized energy states analogous to the Landau levels of electrons in magnetic fields. By designing a laser cavity to induce frequency degeneracy of structured light modes, discrete photonic energy levels with constant spacing can be achieved. This provides a simple yet powerful platform to explore exotic topological physics phenomena using structured light.
Published on 15 May 2024
Group testing for two types of defectives
This paper develops a belief propagation algorithm for efficient group testing when there are two types of defectives, A and B. It constructs pooling designs using finite affine geometry to test items for A, B, and both AB. Simulations evaluate the algorithm's performance for identifying defectives, suggesting effective pooling designs and the impact of short cycles on screening power.
Published on 15 May 2024
Evaluating large language models for explainable fact checking of health claims
This paper evaluates several state-of-the-art large language models on their ability to verify the truth of public health claims and generate natural language explanations justifying their veracity assessments. The study examines zero-shot, few-shot, and fine-tuned approaches across both commercial models like GPT-3.5 and GPT-4 as well as open source models. It uses automatic metrics and human evaluation to analyze performance on claim verification and explanation tasks individually and jointly. Results show GPT-4 leads in zero-shot but with...
Published on 15 May 2024
Collective learning of geometries
This paper proposes using systems of matrix differential equations that model collective motions of abstract particles to perform machine learning on non-Euclidean datasets. These particle swarm models can encode mappings between manifolds and learn transformations and symmetries, like rotations and actions of groups. The paper also discusses statistical models like probability distributions that arise from these particle swarm models and can be used for modeling, inference, and learning over curved geometries.
Published on 15 May 2024
Modeling delta-plutonium with a hybrid density functional
This paper explores using a hybrid density functional called HSE to model the challenging delta phase of plutonium. By tuning the percentage of exact Hartree-Fock exchange in HSE to 7.5%, the authors are able to accurately reproduce experimental measurements of lattice spacing and elastic constants in delta-Pu. They explain this success through an 'Anderson impurity' model where tuning the functional creates selective bonding of 5f electrons, with one involved in pi-bonding while others localize, explaining softening of certain elastic modes...