id stringlengths 16 29 | text stringlengths 86 3.49k | source stringlengths 14 112 |
|---|---|---|
arxiv_dataset-196002311.17033 | Harmonic Functions on Four Dimensions
math.CV math.FA
This paper develops theory for a newly-defined bicomplex hyperbolic harmonic
function with four real-dimensional inputs, in a way that generalizes the
connection between real harmonic functions with two real-dimensional inputs and
complex analytic functions. For example, every bicomplex hyperbolic harmonic
function appears as this paper's newly-defined hyperbolic real part of a
bicomplex analytic function, just as every real harmonic function with two
real-dimensional inputs is the real part of a complex analytic function. In
addition, this connection produces a unique (up to additive constant) and
newly-defined hyperbolic harmonic conjugate function, just as every real
harmonic function has a unique (up to additive constant) real harmonic
conjugate. Finally, the paper determines a bicomplex Poisson kernel function
that produces a corresponding integral representation for bicomplex harmonic
functions, one that generalizes the complex harmonic function Poisson integral
representation.
| arxiv topic:math.CV math.FA |
arxiv_dataset-196012311.17133 | Deployment of a Robust and Explainable Mortality Prediction Model: The
COVID-19 Pandemic and Beyond
cs.LG cs.AI
This study investigated the performance, explainability, and robustness of
deployed artificial intelligence (AI) models in predicting mortality during the
COVID-19 pandemic and beyond. The first study of its kind, we found that
Bayesian Neural Networks (BNNs) and intelligent training techniques allowed our
models to maintain performance amidst significant data shifts. Our results
emphasize the importance of developing robust AI models capable of matching or
surpassing clinician predictions, even under challenging conditions. Our
exploration of model explainability revealed that stochastic models generate
more diverse and personalized explanations thereby highlighting the need for AI
models that provide detailed and individualized insights in real-world clinical
settings. Furthermore, we underscored the importance of quantifying uncertainty
in AI models which enables clinicians to make better-informed decisions based
on reliable predictions. Our study advocates for prioritizing implementation
science in AI research for healthcare and ensuring that AI solutions are
practical, beneficial, and sustainable in real-world clinical environments. By
addressing unique challenges and complexities in healthcare settings,
researchers can develop AI models that effectively improve clinical practice
and patient outcomes.
| arxiv topic:cs.LG cs.AI |
arxiv_dataset-196022311.17233 | Quantifying the redundancy between prosody and text
cs.CL cs.AI cs.IT cs.LG math.IT
Prosody -- the suprasegmental component of speech, including pitch, loudness,
and tempo -- carries critical aspects of meaning. However, the relationship
between the information conveyed by prosody vs. by the words themselves remains
poorly understood. We use large language models (LLMs) to estimate how much
information is redundant between prosody and the words themselves. Using a
large spoken corpus of English audiobooks, we extract prosodic features aligned
to individual words and test how well they can be predicted from LLM
embeddings, compared to non-contextual word embeddings. We find a high degree
of redundancy between the information carried by the words and prosodic
information across several prosodic features, including intensity, duration,
pauses, and pitch contours. Furthermore, a word's prosodic information is
redundant with both the word itself and the context preceding as well as
following it. Still, we observe that prosodic features can not be fully
predicted from text, suggesting that prosody carries information above and
beyond the words. Along with this paper, we release a general-purpose data
processing pipeline for quantifying the relationship between linguistic
information and extra-linguistic features.
| arxiv topic:cs.CL cs.AI cs.IT cs.LG math.IT |
arxiv_dataset-196032311.17333 | Path integral molecular dynamics approximations of quantum canonical
observables
math.NA cs.NA physics.comp-ph
Mean-field molecular dynamics based on path integrals is used to approximate
canonical quantum observables for particle systems consisting of nuclei and
electrons. A computational bottleneck is the sampling from the Gibbs density of
the electron operator, which due to the fermion sign problem has a
computational complexity that scales exponentially with the number of
electrons. In this work we construct an algorithm that approximates the
mean-field Hamiltonian by path integrals for fermions. The algorithm is based
on the determinant of a matrix with components based on Brownian bridges
connecting permuted electron coordinates. The computational work for $n$
electrons is $\mathcal O(n^3)$, which reduces the computational complexity
associated with the fermion sign problem. We analyze a bias resulting from this
approximation and provide a computational error indicator. It remains to
rigorously explain the surprisingly high accuracy.
| arxiv topic:math.NA cs.NA physics.comp-ph |
arxiv_dataset-196042311.17433 | Cospectrality results for signed graphs with two eigenvalues unequal to
$\pm 1$
math.CO
Recently the collection $\cal G$ of all signed graphs for which the adjacency
matrix has all but at most two eigenvalues equal to $\pm 1$ has been
determined. Here we investigate $\cal G$ for cospectral pairs, and for signed
graphs determined by their spectrum (up to switching). If the order is at most
20, the outcome is presented in a clear table. If the spectrum is symmetric we
find all signed graphs in $\cal G$ determined by their spectrum, and we obtain
all signed graphs cospectral with the bipartite double of the complete graph.
In addition we determine all signed graphs cospectral with the Friendship graph
$F_\ell$, and show that there is no connected signed graph cospectral but not
switching equivalent with $F_\ell$.
| arxiv topic:math.CO |
arxiv_dataset-196052311.17533 | A general model and toolkit for the ionization of three or more
electrons in strongly driven molecules using an effective Coulomb potential
for the interaction between bound electrons
physics.atom-ph
We formulate a general three-dimensional semiclassical model for the study of
correlated multielectron escape during fragmentation of molecules driven by
intense infrared laser pulses, while fully accounting for the magnetic field of
the laser pulse. We do so in the context of triple ionization of strongly
driven HeH$_{2}^{+}$. Our model fully accounts for the singularity in the
Coulomb potentials of a recolliding electron with the core and a bound electron
with the core as well as for the interaction of a recolliding with a bound
electron. To avoid artificial autoionization, our model employs effective
potentials to treat the interaction between bound electrons. We focus on triple
and double ionization as well as frustrated triple and frustrated double
ionization. In these processes, we identify and explain the main features of
the sum of the kinetic energies of the final ion fragments.
We find that frustrated double ionization is a major ionization process, and
we identify the different channels and hence different final fragments that are
obtained through frustrated double ionization. Also, we discuss the differences
between frustrated double and triple ionization.
| arxiv topic:physics.atom-ph |
arxiv_dataset-196062311.17633 | Introduction to Transformers: an NLP Perspective
cs.CL cs.AI cs.LG
Transformers have dominated empirical machine learning models of natural
language processing. In this paper, we introduce basic concepts of Transformers
and present key techniques that form the recent advances of these models. This
includes a description of the standard Transformer architecture, a series of
model refinements, and common applications. Given that Transformers and related
deep learning techniques might be evolving in ways we have never seen, we
cannot dive into all the model details or cover all the technical areas.
Instead, we focus on just those concepts that are helpful for gaining a good
understanding of Transformers and their variants. We also summarize the key
ideas that impact this field, thereby yielding some insights into the strengths
and limitations of these models.
| arxiv topic:cs.CL cs.AI cs.LG |
arxiv_dataset-196072311.17733 | Stable Invariants and Their Role in Word Measures on Groups
math.GR math.GT math.PR math.RT
Every word in a free group induces a word measure -- a probability measure
defined via the word map -- on every compact group. This paper presents a
conjectural picture about the role of a plethora of stable invariants of words
in word measures on groups. These invariants generalize the stable commutator
length and include, among others, two invariants recently defined by Wilton:
the stable primitivity rank and a non-oriented analog of stable commutator
length we call stable square length. The conjectures say, roughly, that these
stable invariants control the asymptotics of the expected values of stable
characters, under word measures. We reinforce these conjectures by proving a
version for word measures on wreath products, and by introducing a related
formula for stable irreducible characters of the symmetric group.
| arxiv topic:math.GR math.GT math.PR math.RT |
arxiv_dataset-196082311.17833 | DiG-IN: Diffusion Guidance for Investigating Networks -- Uncovering
Classifier Differences Neuron Visualisations and Visual Counterfactual
Explanations
cs.CV cs.AI cs.LG
While deep learning has led to huge progress in complex image classification
tasks like ImageNet, unexpected failure modes, e.g. via spurious features, call
into question how reliably these classifiers work in the wild. Furthermore, for
safety-critical tasks the black-box nature of their decisions is problematic,
and explanations or at least methods which make decisions plausible are needed
urgently. In this paper, we address these problems by generating images that
optimize a classifier-derived objective using a framework for guided image
generation. We analyze the decisions of image classifiers by visual
counterfactual explanations (VCEs), detection of systematic mistakes by
analyzing images where classifiers maximally disagree, and visualization of
neurons and spurious features. In this way, we validate existing observations,
e.g. the shape bias of adversarially robust models, as well as novel failure
modes, e.g. systematic errors of zero-shot CLIP classifiers. Moreover, our VCEs
outperform previous work while being more versatile.
| arxiv topic:cs.CV cs.AI cs.LG |
arxiv_dataset-196092311.17933 | Arbitrary Controlled Re-Orientation of a Spinning Body by Evolving its
Tensor of Inertia
nlin.CD math-ph math.MP
Bodies with the nonspherical tensor of inertia exhibit a variety of
rotational motion patterns, including chaotic motion, stable periodic
(quasi-periodic) rotation, unstable rotation around the direction close to the
body's second principal axis, featuring a well-known tennis-racket (also known
as Garriott-Dzhanibekov) effect -- series of seemingly spontaneous 180 degrees
flips. These patterns are even more complex if the body's tensor of inertia
(TOI) is changing with time. Changing a body's TOI has been discussed recently
as a tool to perform controllable Garriott-Dzhanibekov flips and similar
maneuvers. In this work, the optimal control of the TOI of the body
(spacecraft, or any other device that admits free rotation in three dimensions)
is used as a means to perform desirable re-orientations of a body with respect
to its angular velocity. Using the spherical TOI as the initial and final point
of the maneuver, we optimize the parameters of the maneuver to achieve and
stabilize the desired orientation of the body's principal axes with respect to
spin angular velocity. It appears that such a procedure allows for finding
arbitrarily complex maneuver trajectories of a spinning body. In particular,
intermediate axis instability can be used to break the alignment of the body's
principal axis and the axis of rotation. Such maneuvers do not require
utilization of propellants and could be straightforwardly used for attitude
control of a spin-stabilized spacecraft. The capabilities of such a method of
angular maneuvering are demonstrated in numerical simulations.
| arxiv topic:nlin.CD math-ph math.MP |
arxiv_dataset-196102311.18033 | Discovering Heavy Neutral Leptons with the Higgs Boson
hep-ph
We study the dominant signatures that arise in Higgs physics at colliders
when extending the Standard Model (SM) with a Yukawa interaction to heavy
neutral leptons (HNL), while suppressing their mixing to active neutrinos. We
focus on the production of HNLs from Higgs bosons that subsequently decay via
the Higgs to SM fermions to determine the experimental reach at the LHC
detectors and far detectors such as FASER and MATHUSLA. We also determine the
impact of precision Higgs constraints on beyond-SM parameters in this scenario.
| arxiv topic:hep-ph |
arxiv_dataset-196112311.18133 | Energy Gap from Step Structure of the Analytically Inverted Non-Additive
Kinetic Potential
physics.chem-ph
The bandgap constitutes a challenging problem in density functional theory
(DFT) methodologies. It is known that the energy gap values calculated by
common DFT approaches are underestimated. The bandgap was also found to be
related to the derivative discontinuity (DD) of the exchange-correlation
potential in the Kohn-Sham formulation of DFT. Several reports have shown that
DD appears as a step on the potential curve. The step structure is a mandatory
structure for aligning the KS energy levels in the ionization potentials in a
dissociated molecule in both fragments and is a function of electron
localisation. Reproducing the step in the DFT framework gives the charge
transfer process and the correct energy gap and describes the source of
dissociation. This step phenomenon has not yet been studied in the non-additive
kinetic potential $v^{\text{NAD}}[\rho_A,\rho_B](\textbf{r})$, a key quantity
used in embedding theories. While $v^{\text{NAD}}[\rho_A,\rho_B](\textbf{r})$
is known to be difficult to approximate, in this work, we explain how an
accurate energy gap can be produced from the analytically inverted
$v^{\text{NAD}}[\rho_A,\rho_B](\textbf{r})$, even if we use the input densities
calculated by the local and semi-local functionals. We used the precisely
calculated $v^{\text{NAD}}[\rho_A,\rho_B](\textbf{r})$ reported in our previous
publication [Phys. Rev. A 106, 042812 (2022)] to produce the energy gap for
some model systems and report in this work the promising accuracy of our
results through the comparison with the results obtained from one of the most
accurate calculations, OEP theory with the KLI local approximation.
| arxiv topic:physics.chem-ph |
arxiv_dataset-196122311.18233 | Semantic Bound and Multi Types, Revisited
cs.LO
Intersection types are a standard tool in operational and semantical studies
of the lambda calculus. De Carvalho showed how multi types, a quantitative
variant of intersection types providing a handy presentation of the relational
denotational model, allows one to extract precise bounds on the number of
$\beta$-steps and the size of normal forms.
In the last few years, de Carvalho's work has been extended and adapted to a
number of lambda calculi, evaluation strategies, and abstract machines. These
works, however, only adapt the first part of his work, that extracts bounds
from multi type derivations, while never consider the second part, which deals
with extracting bounds from the multi types themselves. The reason is that this
second part is more technical, and requires to reason up to type substitutions.
It is however also the most interesting, because it shows that the bounding
power is inherent to the relational model (which is induced by the types,
without the derivations), independently of its presentation as a type system.
Here we dissect and clarify the second part of de Carvalho's work,
establishing a link with principal multi types, and isolating a key property
independent of type substitutions.
| arxiv topic:cs.LO |
arxiv_dataset-196132311.18333 | Spherical Designs for Function Approximation and Beyond
math.NA cs.NA eess.SP
In this paper, we compare two optimization algorithms using full Hessian and
approximation Hessian to obtain numerical spherical designs through their
variational characterization. Based on the obtained spherical design point
sets, we investigate the approximation of smooth and non-smooth functions by
spherical harmonics with spherical designs. Finally, we use spherical framelets
for denoising Wendland functions as an application, which shows the great
potential of spherical designs in spherical data processing.
| arxiv topic:math.NA cs.NA eess.SP |
arxiv_dataset-196142311.18433 | E2PNet: Event to Point Cloud Registration with Spatio-Temporal
Representation Learning
cs.CV
Event cameras have emerged as a promising vision sensor in recent years due
to their unparalleled temporal resolution and dynamic range. While registration
of 2D RGB images to 3D point clouds is a long-standing problem in computer
vision, no prior work studies 2D-3D registration for event cameras. To this
end, we propose E2PNet, the first learning-based method for event-to-point
cloud registration. The core of E2PNet is a novel feature representation
network called Event-Points-to-Tensor (EP2T), which encodes event data into a
2D grid-shaped feature tensor. This grid-shaped feature enables matured
RGB-based frameworks to be easily used for event-to-point cloud registration,
without changing hyper-parameters and the training procedure. EP2T treats the
event input as spatio-temporal point clouds. Unlike standard 3D learning
architectures that treat all dimensions of point clouds equally, the novel
sampling and information aggregation modules in EP2T are designed to handle the
inhomogeneity of the spatial and temporal dimensions. Experiments on the MVSEC
and VECtor datasets demonstrate the superiority of E2PNet over hand-crafted and
other learning-based methods. Compared to RGB-based registration, E2PNet is
more robust to extreme illumination or fast motion due to the use of event
data. Beyond 2D-3D registration, we also show the potential of EP2T for other
vision tasks such as flow estimation, event-to-image reconstruction and object
recognition. The source code can be found at:
https://github.com/Xmu-qcj/E2PNet.
| arxiv topic:cs.CV |
arxiv_dataset-196152311.18533 | A knowledge-driven framework for synthesizing designs from modular
components
cs.RO cs.SE
Creating a design from modular components necessitates three steps: Acquiring
knowledge about available components, conceiving an abstract design concept,
and implementing that concept in a concrete design. The third step entails many
repetitive and menial tasks, such as inserting parts and creating joints
between them. Especially when comparing and implementing design alternatives,
this issue is compounded. We propose a use-case agnostic knowledge-driven
framework to automate the implementation step. In particular, the framework
catalogues the acquired knowledge and the design concept, as well as utilizes
Combinatory Logic Synthesis to synthesize concrete design alternatives. This
minimizes the effort required to create designs, allowing the design space to
be thoroughly explored. We implemented the framework as a plugin for the CAD
software Autodesk Fusion 360. We conducted a case study in which robotic arms
were synthesized from a set of 28 modular components. Based on the case study,
the applicability of the framework is analyzed and discussed.
| arxiv topic:cs.RO cs.SE |
arxiv_dataset-196162311.18633 | The joint spectral radius is pointwise H\"older continuous
math.DS math.OC math.SP
We show that the joint spectral radius is pointwise H\"older continuous. In
addition, the joint spectral radius is locally H\"older continuous for
$\varepsilon$-inflations. In the two-dimensional case, local H\"older
continuity holds on the matrix sets with positive joint spectral radius.
| arxiv topic:math.DS math.OC math.SP |
arxiv_dataset-196172311.18733 | Matrix product state fixed points of non-Hermitian transfer matrices
cond-mat.stat-mech cond-mat.str-el quant-ph
The contraction of tensor networks is a central task in the application of
tensor network methods to the study of quantum and classical many body systems.
In this paper, we investigate the impact of gauge degrees of freedom in the
virtual indices of the tensor network on the contraction process, specifically
focusing on boundary matrix product state methods for contracting
two-dimensional tensor networks. We show that the gauge transformation can
affect the entanglement structures of the eigenstates of the transfer matrix
and change how the physical information is encoded in the eigenstates, which
can influence the accuracy of the numerical simulation. We demonstrate this
effect by looking at two different examples. First, we focus on the local gauge
transformation, and analyze its effect by viewing it as an imaginary-time
evolution governed by a diagonal Hamiltonian. As a specific example, we perform
a numerical analysis in the classical Ising model on the square lattice.
Second, we go beyond the scope of local gauge transformations and study the
antiferromagnetic Ising model on the triangular lattice. The partition function
of this model has two tensor network representations connected by a non-local
gauge transformation, resulting in distinct numerical performances in the
boundary matrix product state calculation.
| arxiv topic:cond-mat.stat-mech cond-mat.str-el quant-ph |
arxiv_dataset-196182311.18833 | VLA FRAMEx. I. Wideband Radio Properties of the AGN in NGC 4388
astro-ph.GA
We present the first results from Karl G. Jansky Very Large Array (VLA)
observations as a part of the Fundamental Reference Active Galactic Nucleus
(AGN) Monitoring Experiment (FRAMEx), a program to understand the relationship
between AGN accretion physics and wavelength-dependent position as a function
of time. With this VLA survey, we investigate the radio properties from a
volume-complete sample of 25 hard X-ray-selected AGNs using the VLA in its
wideband mode. We observed the targets in the A-array configuration at $4-12$
GHz with all polarization products. In this work, we introduce our calibration
and imaging methods for this survey, and we present our results and analysis
for the radio quiet AGN NGC 4388. We calibrated and imaged these data using the
multi-term, multi-frequency synthesis imaging algorithm to determine its
spatial, spectral and polarization structure across a continuous $4-12$ GHz
band. In the AGN, we measure a broken power law spectrum with $\alpha=-0.06$
below a break frequency of 7.3 GHz and $\alpha=-0.34$ above. We detect
polarization at sub-arcsecond resolution across both the AGN and a secondary
radio knot. We compare our results to ancillary data and find that the VLA
radio continuum is likely due to AGN winds interacting with the local
interstellar medium that gets resolved away at sub-parsec spatial scales as
probed by the Very Long Baseline Array. A well-known ionization cone to the
southwest of the AGN appears likely to be projected material onto the underside
of the disk of the host galaxy.
| arxiv topic:astro-ph.GA |
arxiv_dataset-196192312.00092 | Mixture of Gaussian-distributed Prototypes with Generative Modelling for
Interpretable and Trustworthy Image Recognition
cs.CV
Prototypical-part methods, e.g., ProtoPNet, enhance interpretability in image
recognition by linking predictions to training prototypes, thereby offering
intuitive insights into their decision-making. Existing methods, which rely on
a point-based learning of prototypes, typically face two critical issues: 1)
the learned prototypes have limited representation power and are not suitable
to detect Out-of-Distribution (OoD) inputs, reducing their decision
trustworthiness; and 2) the necessary projection of the learned prototypes back
into the space of training images causes a drastic degradation in the
predictive performance. Furthermore, current prototype learning adopts an
aggressive approach that considers only the most active object parts during
training, while overlooking sub-salient object regions which still hold crucial
classification information. In this paper, we present a new generative paradigm
to learn prototype distributions, termed as Mixture of Gaussian-distributed
Prototypes (MGProto). The distribution of prototypes from MGProto enables both
interpretable image classification and trustworthy recognition of OoD inputs.
The optimisation of MGProto naturally projects the learned prototype
distributions back into the training image space, thereby addressing the
performance degradation caused by prototype projection. Additionally, we
develop a novel and effective prototype mining strategy that considers not only
the most active but also sub-salient object parts. To promote model
compactness, we further propose to prune MGProto by removing prototypes with
low importance priors. Experiments on CUB-200-2011, Stanford Cars, Stanford
Dogs, and Oxford-IIIT Pets datasets show that MGProto achieves state-of-the-art
image recognition and OoD detection performances, while providing encouraging
interpretability results.
| arxiv topic:cs.CV |
arxiv_dataset-196202312.00192 | Benchmarking and Enhancing Disentanglement in Concept-Residual Models
cs.LG cs.CV
Concept bottleneck models (CBMs) are interpretable models that first predict
a set of semantically meaningful features, i.e., concepts, from observations
that are subsequently used to condition a downstream task. However, the model's
performance strongly depends on the engineered features and can severely suffer
from incomplete sets of concepts. Prior works have proposed a side channel -- a
residual -- that allows for unconstrained information flow to the downstream
task, thus improving model performance but simultaneously introducing
information leakage, which is undesirable for interpretability. This work
proposes three novel approaches to mitigate information leakage by
disentangling concepts and residuals, investigating the critical balance
between model performance and interpretability. Through extensive empirical
analysis on the CUB, OAI, and CIFAR 100 datasets, we assess the performance of
each disentanglement method and provide insights into when they work best.
Further, we show how each method impacts the ability to intervene over the
concepts and their subsequent impact on task performance.
| arxiv topic:cs.LG cs.CV |
arxiv_dataset-196212312.00292 | SEPSIS: I Can Catch Your Lies -- A New Paradigm for Deception Detection
cs.CL
Deception is the intentional practice of twisting information. It is a
nuanced societal practice deeply intertwined with human societal evolution,
characterized by a multitude of facets. This research explores the problem of
deception through the lens of psychology, employing a framework that
categorizes deception into three forms: lies of omission, lies of commission,
and lies of influence. The primary focus of this study is specifically on
investigating only lies of omission. We propose a novel framework for deception
detection leveraging NLP techniques. We curated an annotated dataset of 876,784
samples by amalgamating a popular large-scale fake news dataset and scraped
news headlines from the Twitter handle of Times of India, a well-known Indian
news media house. Each sample has been labeled with four layers, namely: (i)
the type of omission (speculation, bias, distortion, sounds factual, and
opinion), (ii) colors of lies(black, white, etc), and (iii) the intention of
such lies (to influence, etc) (iv) topic of lies (political, educational,
religious, etc). We present a novel multi-task learning pipeline that leverages
the dataless merging of fine-tuned language models to address the deception
detection task mentioned earlier. Our proposed model achieved an F1 score of
0.87, demonstrating strong performance across all layers including the type,
color, intent, and topic aspects of deceptive content. Finally, our research
explores the relationship between lies of omission and propaganda techniques.
To accomplish this, we conducted an in-depth analysis, uncovering compelling
findings. For instance, our analysis revealed a significant correlation between
loaded language and opinion, shedding light on their interconnectedness. To
encourage further research in this field, we will be making the models and
dataset available with the MIT License, making it favorable for open-source
research.
| arxiv topic:cs.CL |
arxiv_dataset-196222312.00392 | Study and Survey on Gesture Recognition Systems
cs.CV
In recent years, there has been a considerable amount of research in the
Gesture Recognition domain, mainly owing to the technological advancements in
Computer Vision. Various new applications have been conceptualised and
developed in this field. This paper discusses the implementation of gesture
recognition systems in multiple sectors such as gaming, healthcare, home
appliances, industrial robots, and virtual reality. Different methodologies for
capturing gestures are compared and contrasted throughout this survey. Various
data sources and data acquisition techniques have been discussed. The role of
gestures in sign language has been studied and existing approaches have been
reviewed. Common challenges faced while building gesture recognition systems
have also been explored.
| arxiv topic:cs.CV |
arxiv_dataset-196232312.00492 | Generic multi-particle transverse momentum correlations as a new tool
for studying nuclear structure at the energy frontier
nucl-th nucl-ex
The mean transverse momentum of produced particles, [pt], and its
event-by-event fluctuations give direct access to the initial conditions of
ultra-relativistic heavy-ion collisions and help probe the colliding nuclei's
structure. The [pt] fluctuations can be studied via multi-particle pt
correlations; so far, only the lowest four orders have been studied.
Higher-order fluctuations can provide stronger constraints on the initial
conditions and improved sensitivity to the detailed nuclear structure; however,
their direct implementation can be challenging and is still lacking. In this
paper, we apply a generic recursive algorithm for the genuine multi-particle pt
correlations, which enables the accurate study of higher-order [pt]
fluctuations without computationally heavy processing for the first time. With
this algorithm, we will examine the power of multi-particle pt correlations
through Monte Carlo model studies with different nuclear structures. The impact
on the nuclear structure studies, including the nuclear deformation and
triaxial structure, will be discussed. These results will demonstrate the
usefulness of multi-particle pt correlations for studying nuclear structure in
high-energy nuclei collisions at RHIC and the LHC, which could serve as
complementary to existing low-energy nuclear structure studies.
| arxiv topic:nucl-th nucl-ex |
arxiv_dataset-196242312.00592 | Tracking Object Positions in Reinforcement Learning: A Metric for
Keypoint Detection (extended version)
cs.LG cs.CV cs.RO
Reinforcement learning (RL) for robot control typically requires a detailed
representation of the environment state, including information about
task-relevant objects not directly measurable. Keypoint detectors, such as
spatial autoencoders (SAEs), are a common approach to extracting a
low-dimensional representation from high-dimensional image data. SAEs aim at
spatial features such as object positions, which are often useful
representations in robotic RL. However, whether an SAE is actually able to
track objects in the scene and thus yields a spatial state representation well
suited for RL tasks has rarely been examined due to a lack of established
metrics. In this paper, we propose to assess the performance of an SAE instance
by measuring how well keypoints track ground truth objects in images. We
present a computationally lightweight metric and use it to evaluate common
baseline SAE architectures on image data from a simulated robot task. We find
that common SAEs differ substantially in their spatial extraction capability.
Furthermore, we validate that SAEs that perform well in our metric achieve
superior performance when used in downstream RL. Thus, our metric is an
effective and lightweight indicator of RL performance before executing
expensive RL training. Building on these insights, we identify three key
modifications of SAE architectures to improve tracking performance.
| arxiv topic:cs.LG cs.CV cs.RO |
arxiv_dataset-196252312.00692 | VisionaryVR: An Optical Simulation Tool for Evaluating and Optimizing
Vision Correction Solutions in Virtual Reality
cs.CV
Developing and evaluating vision science methods require robust and efficient
tools for assessing their performance in various real-world scenarios. This
study presents a novel virtual reality (VR) simulation tool that simulates
real-world optical methods while giving high experimental control to the
experiment. The tool incorporates an experiment controller, to smoothly and
easily handle multiple conditions, a generic eye-tracking controller, that
works with most common VR eye-trackers, a configurable defocus simulator, and a
generic VR questionnaire loader to assess participants' behavior in virtual
reality. This VR-based simulation tool bridges the gap between theoretical and
applied research on new optical methods, corrections, and therapies. It enables
vision scientists to increase their research tools with a robust, realistic,
and fast research environment.
| arxiv topic:cs.CV |
arxiv_dataset-196262312.00792 | Visualization and Characterization of Agricultural Sprays Using Machine
Learning based Digital Inline Holography
physics.flu-dyn
Accurate characterization of agricultural sprays is crucial to predict in
field performance of liquid applied crop protection products. Here we introduce
a robust and efficient machine learning (ML) based Digital In-line Holography
(DIH) to accurately characterize the droplet field for a wide range of
agricultural spray nozzles. Compared to non-ML methods, our method enhances
accuracy, generalizability, and processing speed. Our approach employs two
neural networks: a modified U-Net to obtain the 3D droplet field from the
numerically reconstructed optical field, followed by a VGG16 classifier to
reduce false positives from the U-Net prediction. The modified U-Net is trained
using holograms generated using a single spray nozzle at three spray locations;
center, half-span, and the spray edge to create training data with various
number densities and droplet size ranges. VGG16 is trained via the minimum
intensity projection of the droplet 3D point spread function. Data augmentation
is used to increase the efficiency of classification and make the algorithm
generalizable for different measurement settings. The model is validated via
NIST traceable glass beads and six agricultural spray nozzles representing
various spray characteristics. The results demonstrate a high accuracy rate,
with over 90% droplet extraction and less than 5% false positives. Compared to
traditional spray measurement techniques, our method offers a significant leap
forward in spatial resolution and generalizability. In particular, our method
can extract the real cumulative volume distribution of the NIST beads, where
the laser diffraction is biased towards droplets moving at slower speeds.
Additionally, the ML-based DIH enables the estimation of mass and momentum flux
at different locations and the calculation of relative velocities of droplet
pairs, which are difficult to obtain via conventional techniques.
| arxiv topic:physics.flu-dyn |
arxiv_dataset-196272312.00892 | Black-Litterman Portfolio Optimization with Noisy Intermediate-Scale
Quantum Computers
quant-ph
In this work, we demonstrate a practical application of noisy
intermediate-scale quantum (NISQ) algorithms to enhance subroutines in the
Black-Litterman (BL) portfolio optimization model. As a proof of concept, we
implement a 12-qubit example for selecting 6 assets out of a 12-asset pool. Our
approach involves predicting investor views with quantum machine learning (QML)
and addressing the subsequent optimization problem using the variational
quantum eigensolver (VQE). The solutions obtained from VQE exhibit a high
approximation ratio behavior, and consistently outperform several common
portfolio models in backtesting over a long period of time. A unique aspect of
our VQE scheme is that after the quantum circuit is optimized, only a minimal
number of samplings is required to give a high approximation ratio result since
the probability distribution should be concentrated on high-quality solutions.
We further emphasize the importance of employing only a small number of final
samplings in our scheme by comparing the cost with those obtained from an
exhaustive search and random sampling. The power of quantum computing can be
anticipated when dealing with a larger-size problem due to the linear growth of
the required qubit resources with the problem size. This is in contrast to
classical computing where the search space grows exponentially with the problem
size and would quickly reach the limit of classical computers.
| arxiv topic:quant-ph |
arxiv_dataset-196282312.00992 | Improving Normative Modeling for Multi-modal Neuroimaging Data using
mixture-of-product-of-experts variational autoencoders
cs.LG
Normative models in neuroimaging learn the brain patterns of healthy
population distribution and estimate how disease subjects like Alzheimer's
Disease (AD) deviate from the norm. Existing variational autoencoder
(VAE)-based normative models using multimodal neuroimaging data aggregate
information from multiple modalities by estimating product or averaging of
unimodal latent posteriors. This can often lead to uninformative joint latent
distributions which affects the estimation of subject-level deviations. In this
work, we addressed the prior limitations by adopting the
Mixture-of-Product-of-Experts (MoPoE) technique which allows better modelling
of the joint latent posterior. Our model labelled subjects as outliers by
calculating deviations from the multimodal latent space. Further, we identified
which latent dimensions and brain regions were associated with abnormal
deviations due to AD pathology.
| arxiv topic:cs.LG |
arxiv_dataset-196292312.01092 | A Semi-Supervised Deep Learning Approach to Dataset Collection for
Query-By-Humming Task
cs.SD cs.LG eess.AS
Query-by-Humming (QbH) is a task that involves finding the most relevant song
based on a hummed or sung fragment. Despite recent successful commercial
solutions, implementing QbH systems remains challenging due to the lack of
high-quality datasets for training machine learning models. In this paper, we
propose a deep learning data collection technique and introduce Covers and
Hummings Aligned Dataset (CHAD), a novel dataset that contains 18 hours of
short music fragments, paired with time-aligned hummed versions. To expand our
dataset, we employ a semi-supervised model training pipeline that leverages the
QbH task as a specialized case of cover song identification (CSI) task.
Starting with a model trained on the initial dataset, we iteratively collect
groups of fragments of cover versions of the same song and retrain the model on
the extended data. Using this pipeline, we collect over 308 hours of additional
music fragments, paired with time-aligned cover versions. The final model is
successfully applied to the QbH task and achieves competitive results on
benchmark datasets. Our study shows that the proposed dataset and training
pipeline can effectively facilitate the implementation of QbH systems.
| arxiv topic:cs.SD cs.LG eess.AS |
arxiv_dataset-196302312.01192 | Jacobian schemes arising from hypersurface arrangements in $\mathbb P^n$
math.AG math.AC
Freeness is an important property of a hypersurface arrangement, although its
presence is not well understood. A hypersurface arrangement in $\PP^n$ is free
if $S/J$ is Cohen-Macaulay (CM), where $S = K[x_0,\ldots,x_n]$ and $J$ is the
Jacobian ideal. We study three related unmixed ideals: $J^{top}$, the
intersection of height two primary components, $\sqrt{J^{top}}$, the radical of
$J^{top}$, and when the $f_i$ are smooth we also study $\sqrt{J}$. Under mild
hypotheses, we show that these ideals are CM. This establishes a full
generalization of an earlier result with Schenck from hyperplane arrangements
to hypersurface arrangements. If the hypotheses fail for an arrangement in
projective $3$-space, the Hartshorne-Rao module measures the failure of CMness.
We establish consequences for the even liaison classes of $J^{top}$ and
$\sqrt{J}$.
| arxiv topic:math.AG math.AC |
arxiv_dataset-196312312.01292 | Joint Beam Scheduling and Power Optimization for Beam Hopping LEO
Satellite Systems
cs.NI eess.SP
Low earth orbit (LEO) satellite communications can provide ubiquitous and
reliable services, making it an essential part of the Internet of Everything
network. Beam hopping (BH) is an emerging technology for effectively addressing
the issue of low resource utilization caused by the non-uniform spatio-temporal
distribution of traffic demands. However, how to allocate multi-dimensional
resources in a timely and efficient way for the highly dynamic LEO satellite
systems remains a challenge. This paper proposes a joint beam scheduling and
power optimization beam hopping (JBSPO-BH) algorithm considering the
differences in the geographic distribution of sink nodes. The JBSPO-BH
algorithm decouples the original problem into two sub-problems. The beam
scheduling problem is modelled as a potential game, and the Nash equilibrium
(NE) point is obtained as the beam scheduling strategy. Moreover, the penalty
function interior point method is applied to optimize the power allocation.
Simulation results show that the JBSPO-BH algorithm has low time complexity and
fast convergence and achieves better performance both in throughput and
fairness. Compared with greedy-based BH, greedy-based BH with the power
optimization, round-robin BH, Max-SINR BH and satellite resource allocation
algorithm, the throughput of the proposed algorithm is improved by 44.99%,
20.79%, 156.06%, 15.39% and 8.17%, respectively.
| arxiv topic:cs.NI eess.SP |
arxiv_dataset-196322312.01392 | Neural Network Characterization and Entropy Regulated Data Balancing
through Principal Component Analysis
cs.LG
This paper examines the relationship between the behavior of a neural network
and the distribution formed from the projections of the data records into the
space spanned by the low-order principal components of the training data. For
example, in a benchmark calculation involving rotated and unrotated MNIST
digits, classes (digits) that are mapped far from the origin in a
low-dimensional principal component space and that overlap minimally with other
digits converge rapidly and exhibit high degrees of accuracy in neural network
calculations that employ the associated components of each data record as
inputs. Further, if the space spanned by these low-order principal components
is divided into bins and the input data records that are mapped into a given
bin averaged, the resulting pattern can be distinguished by its geometric
features which interpolate between those of adjacent bins in an analogous
manner to variational autoencoders. Based on this observation, a simply
realized data balancing procedure can be realized by evaluating the entropy
associated with each histogram bin and subsequently repeating the original
image data associated with the bin by a number of times that is determined from
this entropy.
| arxiv topic:cs.LG |
arxiv_dataset-196332312.01492 | The Multilinear Rank and Core of Trifocal Grassmann Tensors
math.AG
Closed formulas for the multilinear rank of trifocal Grassmann tensors are
obtained. An alternative process to the standard HOSVD is introduced for the
computation of the core of trifocal Grassmann tensors. Both of these results
are obtained, under natural genericity conditions, leveraging the canonical
form for these tensors, obtained by the same authors in a previous work. A
gallery of explicit examples is also included.
| arxiv topic:math.AG |
arxiv_dataset-196342312.01592 | Expand BERT Representation with Visual Information via Grounded Language
Learning with Multimodal Partial Alignment
cs.CL
Language models have been supervised with both language-only objective and
visual grounding in existing studies of visual-grounded language learning.
However, due to differences in the distribution and scale of visual-grounded
datasets and language corpora, the language model tends to mix up the context
of the tokens that occurred in the grounded data with those that do not. As a
result, during representation learning, there is a mismatch between the visual
information and the contextual meaning of the sentence. To overcome this
limitation, we propose GroundedBERT - a grounded language learning method that
enhances the BERT representation with visually grounded information.
GroundedBERT comprises two components: (i) the original BERT which captures the
contextual representation of words learned from the language corpora, and (ii)
a visual grounding module which captures visual information learned from
visual-grounded datasets. Moreover, we employ Optimal Transport (OT),
specifically its partial variant, to solve the fractional alignment problem
between the two modalities. Our proposed method significantly outperforms the
baseline language models on various language tasks of the GLUE and SQuAD
datasets.
| arxiv topic:cs.CL |
arxiv_dataset-196352312.01692 | Risk-Controlling Model Selection via Guided Bayesian Optimization
cs.LG cs.AI stat.ME stat.ML
Adjustable hyperparameters of machine learning models typically impact
various key trade-offs such as accuracy, fairness, robustness, or inference
cost. Our goal in this paper is to find a configuration that adheres to
user-specified limits on certain risks while being useful with respect to other
conflicting metrics. We solve this by combining Bayesian Optimization (BO) with
rigorous risk-controlling procedures, where our core idea is to steer BO
towards an efficient testing strategy. Our BO method identifies a set of Pareto
optimal configurations residing in a designated region of interest. The
resulting candidates are statistically verified and the best-performing
configuration is selected with guaranteed risk levels. We demonstrate the
effectiveness of our approach on a range of tasks with multiple desiderata,
including low error rates, equitable predictions, handling spurious
correlations, managing rate and distortion in generative models, and reducing
computational costs.
| arxiv topic:cs.LG cs.AI stat.ME stat.ML |
arxiv_dataset-196362312.01792 | Wild-Tab: A Benchmark For Out-Of-Distribution Generalization In Tabular
Regression
cs.LG
Out-of-Distribution (OOD) generalization, a cornerstone for building robust
machine learning models capable of handling data diverging from the training
set's distribution, is an ongoing challenge in deep learning. While significant
progress has been observed in computer vision and natural language processing,
its exploration in tabular data, ubiquitous in many industrial applications,
remains nascent. To bridge this gap, we present Wild-Tab, a large-scale
benchmark tailored for OOD generalization in tabular regression tasks. The
benchmark incorporates 3 industrial datasets sourced from fields like weather
prediction and power consumption estimation, providing a challenging testbed
for evaluating OOD performance under real-world conditions. Our extensive
experiments, evaluating 10 distinct OOD generalization methods on Wild-Tab,
reveal nuanced insights. We observe that many of these methods often struggle
to maintain high-performance levels on unseen data, with OOD performance
showing a marked drop compared to in-distribution performance. At the same
time, Empirical Risk Minimization (ERM), despite its simplicity, delivers
robust performance across all evaluations, rivaling the results of
state-of-the-art methods. Looking forward, we hope that the release of Wild-Tab
will facilitate further research on OOD generalization and aid in the
deployment of machine learning models in various real-world contexts where
handling distribution shifts is a crucial requirement.
| arxiv topic:cs.LG |
arxiv_dataset-196372312.01892 | PSR J0210+5845; An ultra wide binary pulsar with a B6V main-sequence
star companion
astro-ph.HE
We report on radio timing observations of PSR J0210+5845 which reveal large
deviations from typical pulsar spin-down behaviour. We interpret these
deviations as being due to binary motion around the $V=13.5$ star 2MASS
J02105640$+$5845176, which is coincident in celestial position and distance
with the pulsar. Archival observations and new optical spectroscopy identify
this star as a B6V star with a temperature of $T_\mathrm{eff}\approx 14\,000$K
and a mass of $M_\mathrm{c}= 3.5$ to $3.8$M$_\odot$, making it the lowest mass
main-sequence star known orbiting a non-recycled pulsar. We found that the
timing observations constrain the binary orbit to be wide and moderately
eccentric, with an orbital period of $P_\mathrm{b}=47^{+40}_{-14}$yr and
eccentricity $e=0.46^{+0.10}_{-0.07}$. We predict that the next periastron
passage will occur between 2030 and 2034. Due to the low companion mass, we
find that the probability for a system with the properties of PSR J0210+5845
and its binary companion to survive the supernova is low. We show that a low
velocity and fortuitously directed natal kick is required for the binary to
remain bound during the supernova explosion, and argue that an electron-capture
supernova is a plausible formation scenario for the pulsar.
| arxiv topic:astro-ph.HE |
arxiv_dataset-196382312.01992 | Whence Nonlocality? Removing spooky action at a distance from the de
Broglie Bohm pilot-wave theory using a time-symmetric version of de Broglie
double solution
quant-ph physics.hist-ph
In this work, we review and extend a version of the old attempt made by Louis
de broglie for interpreting quantum mechanics in realistic terms, namely the
double solution. In this theory quantum particles are localized waves, i.e,
solitons, that are solutions of relativistic nonlinear field equations. The
theory that we present here is the natural extension of this old work and
relies on a strong time-symmetry requiring the presence of advanced and
retarded waves converging on particles. Using this method, we are able to
justify wave-particle duality and to explain the violations of Bell's
inequalities. Moreover, the theory recovers the predictions of the pilot-wave
theory of de Borglie and Bohm, often known as Bohmian mechanics. As a direct
consequence, we reinterpret the nonlocal action at a distance presents in the
pilot-wave theory. In the double solution developed here there is fundamentally
no action at a distance but the theory requires a form of superdeterminism
driven by time-symmetry.
| arxiv topic:quant-ph physics.hist-ph |
arxiv_dataset-196392312.02092 | Robust Detrending of Spatially Correlated Systematics in Kepler Light
Curves Using Low-Rank Methods
astro-ph.EP astro-ph.IM
Light curves produced by wide-field exoplanet transit surveys such as CoRoT,
Kepler, and TESS are affected by sensor-wide systematic noise which is
correlated both spatiotemporally and with other instrumental parameters such as
photometric magnitude. Robust and effective systematics mitigation is necessary
to achieve the level of photometric accuracy required to detect exoplanet
transits and to faithfully recover other forms of intrinsic astrophysical
variability. We demonstrate the feasibility of a new exploratory algorithm to
remove spatially-correlated systematic noise and detrend light curves obtained
from wide-field transit surveys. This spatial systematics algorithm is
data-driven and fits a low-rank linear model for the systematics conditioned on
a total-variation spatial constraint. The total-variation constraint models
spatial systematic structure across the sensor on a foundational level. The fit
is performed using gradient descent applied to, a variable reduced
least-squares penalty and a modified form of total-variation prior; both the
systematics basis vectors and their weighting coefficients are iteratively
varied. The algorithm was numerically evaluated against a reference principal
component analysis, using both signal injection on a selected Kepler dataset,
as well as full simulations within the same Kepler coordinate framework. We
find our algorithm to reduce overfitting of astrophysical variability over
longer signal timescales (days) while performing comparably relative to the
reference method for exoplanet transit timescales. The algorithm performance
and application is assessed and future development outlined.
| arxiv topic:astro-ph.EP astro-ph.IM |
arxiv_dataset-196402312.02192 | DiverseDream: Diverse Text-to-3D Synthesis with Augmented Text Embedding
cs.CV
Text-to-3D synthesis has recently emerged as a new approach to sampling 3D
models by adopting pretrained text-to-image models as guiding visual priors. An
intriguing but underexplored problem with existing text-to-3D methods is that
3D models obtained from the sampling-by-optimization procedure tend to have
mode collapses, and hence poor diversity in their results. In this paper, we
provide an analysis and identify potential causes of such a limited diversity,
which motivates us to devise a new method that considers the joint generation
of different 3D models from the same text prompt. We propose to use augmented
text prompts via textual inversion of reference images to diversify the joint
generation. We show that our method leads to improved diversity in text-to-3D
synthesis qualitatively and quantitatively. Project page:
https://diversedream.github.io
| arxiv topic:cs.CV |
arxiv_dataset-196412312.02292 | Thermodynamics and decay of de Sitter vacuum
gr-qc cond-mat.other
We discuss the consequences of unique symmetry of de Sitter spacetime, which
is invariant under the modified translations, ${\bf r}\rightarrow {\bf r}
-e^{Ht}{\bf a}$, where $H$ is the Hubble parameter. Due to this symmetry, all
the comoving observers at any point of the de Sitter space perceive the de
Sitter environment as the thermal bath with temperature $T=H/\pi$, which is
twice larger than the Gibbons-Hawking temperature of the cosmological horizon.
This leads to the heat exchange between gravity and matter, and to instability
of de Sitter state towards the creation of matter, its further heating, and
finally to the decay of the de Sitter state. The temperature $T=H/\pi$
determines different processes in the de Sitter environment, which are not
possible in Minkowski vacuum, such as the process of ionization of an atom.
This temperature also determines the local entropy of the de Sitter vacuum
state, and this allows us to calculate the total entropy inside the
cosmological horizon. The result reproduces the Gibbons-Hawking area law, which
is related to the cosmological horizon, $S_{\rm hor}=4\pi KA$, where
$K=1/(16\pi G)$. This supports the holographic properties of the cosmological
event horizon. We extend the consideration of the local thermodynamics of the
de Sitter state using the $f({\cal R})$ gravity. In this thermodynamics, the
Ricci scalar curvature ${\cal R}$ and the effective gravitational coupling $K$
are thermodynamically conjugate variables. The holographic connection between
the bulk entropy of the Hubble volume and the surface entropy of the
cosmological horizon remains the same. Such connection takes place only in the
$3+1$ spacetime, where there is the special symmetry due to which the variables
$K$ and ${\cal R}$ have the same dimensionality. We also consider the lessons
from the de Sitter symmetry for the thermodynamics of black and white holes.
| arxiv topic:gr-qc cond-mat.other |
arxiv_dataset-196422312.02392 | Instance Space Analysis of Search-Based Software Testing
cs.SE
Search-based software testing (SBST) is now a mature area, with numerous
techniques developed to tackle the challenging task of software testing. SBST
techniques have shown promising results and have been successfully applied in
the industry to automatically generate test cases for large and complex
software systems. Their effectiveness, however, is problem-dependent. In this
paper, we revisit the problem of objective performance evaluation of SBST
techniques considering recent methodological advances -- in the form of
Instance Space Analysis (ISA) -- enabling the strengths and weaknesses of SBST
techniques to be visualized and assessed across the broadest possible space of
problem instances (software classes) from common benchmark datasets. We
identify features of SBST problems that explain why a particular instance is
hard for an SBST technique, reveal areas of hard and easy problems in the
instance space of existing benchmark datasets, and identify the strengths and
weaknesses of state-of-the-art SBST techniques. In addition, we examine the
diversity and quality of common benchmark datasets used in experimental
evaluations.
| arxiv topic:cs.SE |
arxiv_dataset-196432312.02492 | Cosmological reconstruction and $\Lambda$CDM universe in $f(Q,C)$
gravity
gr-qc hep-th
Symmetric Teleparallel Gravity allows for the reformulation of gravity in the
form of nonmetricity by vanishing the contorsion term in the generic affine
connection. Our focus is on investigating a recently proposed extension of this
theory in which the Lagrangian has the form $f(Q,C)$ by incorporating the
boundary term $C$. In this work, we first use a reconstruction approach in
$f(Q,C)$ gravity that might admit the $\Lambda$CDM expansion history.
Furthermore, we perform a novel approach for cosmological reconstruction of
$f(Q,C)$ gravity in terms of e-folding, and it shows how any FLRW cosmology can
arise from a specific $f(Q,C)$ gravity. A variety of instances are provided
using this approach in which $f(Q, C)$ gravity is reconstructed to yield the
well-known cosmic evolution: $\Lambda$CDM era, acceleration/deceleration era
which is equivalent to the presence of phantom and non-phantom matter,
late-time acceleration with the crossing of phantom-divide line and transient
phantom era.
| arxiv topic:gr-qc hep-th |
arxiv_dataset-196442312.02592 | FRAPPE: A Group Fairness Framework for Post-Processing Everything
cs.LG cs.CY
Despite achieving promising fairness-error trade-offs, in-processing
mitigation techniques for group fairness cannot be employed in numerous
practical applications with limited computation resources or no access to the
training pipeline of the prediction model. In these situations, post-processing
is a viable alternative. However, current methods are tailored to specific
problem settings and fairness definitions and hence, are not as broadly
applicable as in-processing. In this work, we propose a framework that turns
any regularized in-processing method into a post-processing approach. This
procedure prescribes a way to obtain post-processing techniques for a much
broader range of problem settings than the prior post-processing literature. We
show theoretically and through extensive experiments that our framework
preserves the good fairness-error trade-offs achieved with in-processing and
can improve over the effectiveness of prior post-processing methods. Finally,
we demonstrate several advantages of a modular mitigation strategy that
disentangles the training of the prediction model from the fairness mitigation,
including better performance on tasks with partial group labels.
| arxiv topic:cs.LG cs.CY |
arxiv_dataset-196452312.02692 | Gate-tunable graphene Josephson diode effect due to magnetochiral
anisotropy
cond-mat.supr-con
Usually the magnetochiral anisotropy related Josephson diode effect is
assumed to be based on conventional two-dimensional electron gas, such as the
InAs quantum well. Here we propose a graphene-based Josephson junction as a
broadly gate-tunable platform for achieving nonreciprocal supercurrent within
the context of magnetochiral anisotropy. We show that the resulting
nonreciprocal supercurrents will exhibit a sign reversal when the graphene
switches from $n$-type doping to $p$-type doping. Particularly, the magnitude
of the nonreciprocity is highly sensitive to the electrostatic doping level of
graphene, enabling gate control of the diode efficiency from zero up to
approximately $40\%$. This giant gate-tunability stems from the chiral nature
of the pseudo-relativistic carriers in grapehe, allowing the graphene Josephson
diode emerges as a promising element for advanced superconducting circuits and
computation devices. Moreover, we have also obtained the so-called $0-\pi$-like
phase transitions in the current-phase relation, in coincidence with recent
experimental finding.
| arxiv topic:cond-mat.supr-con |
arxiv_dataset-196462312.02792 | Hydrodynamic equations for space-inhomogeneous aggregating fluids with
first-principle kinetic coefficients
cond-mat.stat-mech
We derive from the first principles new hydrodynamic equations --
Smoluchowski-Euler equations for aggregation kinetics in space-inhomogeneous
fluids with fluxes. Starting from Boltzmann equations, we obtain microscopic
expressions for aggregation rates for clusters of different sizes and observe
that they significantly differ from currently used phenomenological rates.
Moreover, we show that for a complete description of aggregating systems, novel
kinetic coefficients are needed. They share properties of transport and
reaction-rate coefficients; for them we report microscopic expressions. For two
representative examples -- aggregation of particles at sedimentation and
aggregation after an explosion we numerically solve Smoluchowski-Euler
equations and perform Direct Simulation Monte Carlo (DSMC). We find that while
the new theory agrees well with DSMC results, a noticeable difference is
observed for the phenomenological theory. This manifests the unreliability of
the currently used phenomenological theory and the need to apply new,
first-principle equations.
| arxiv topic:cond-mat.stat-mech |
arxiv_dataset-196472312.02892 | Safe Stabilization with Model Uncertainties: A Universal Formula with
Gaussian Process Learning
eess.SY cs.SY
A combination of control Lyapunov functions (CLFs) and control barrier
functions (CBFs) forms an efficient framework for addressing control challenges
in safe stabilization. In our previous research, we developed an analytical
control strategy, namely the universal formula, that incorporates CLF and CBF
conditions for safe stabilization. However, successful implementation of this
universal formula relies on an accurate model, as any mismatch between the
model and the actual system can compromise stability and safety. In this paper,
we propose a new universal formula that leverages Gaussian processes (GPs)
learning to address safe stabilization in the presence of model uncertainty. By
utilizing the results related to bounded learning errors, we achieve a high
probability of stability and safety guarantees with the proposed universal
formula. Additionally, we introduce a probabilistic compatibility condition to
evaluate conflicts between the modified CLF and CBF conditions with GP learning
results. In cases where compatibility assumptions fail and control system
limits are present, we propose a modified universal formula that relaxes
stability constraints and a projection-based method accommodating control
limits. We illustrate the effectiveness of our approach through a simulation of
adaptive cruise control (ACC), highlighting its potential for practical
applications in real-world scenarios.
| arxiv topic:eess.SY cs.SY |
arxiv_dataset-196482312.02992 | Advancing Web Accessibility -- A guide to transitioning Design Systems
from WCAG 2.0 to WCAG 2.1
cs.HC cs.SE
This research focuses on the critical process of upgrading a Design System
from Web Content Accessibility Guidelines (WCAG) 2.0 to WCAG 2.1, which is an
essential step in enhancing web accessibility. It emphasizes the importance of
staying up to date on increasing accessibility requirements, as well as the
critical function of Design Systems in supporting inclusion in digital
environments. The article lays out a complete strategy for meeting WCAG 2.1
compliance. Assessment, strategic planning, implementation, and testing are all
part of this strategy. The need for collaboration and user involvement is
emphasized as critical strategies and best practices for a successful migration
journey. In addition, the article digs into migration barriers and discusses
significant lessons acquired, offering a realistic view of the intricacies of
this transforming road. Finally, it is a practical guide and a necessary
resource for organizations committed to accessible and user-centered design.
The document provides them with the knowledge and resources they need to
navigate the changing world of web accessibility properly.
| arxiv topic:cs.HC cs.SE |
arxiv_dataset-196492312.03092 | Coloring Groups
math.CO math.GR
We introduce coloring groups, which are permutation groups obtained from a
proper edge coloring of a graph. These groups generalize the generalized toggle
groups of Striker (which themselves generalize the toggle groups introduced by
Cameron and Fon-der-Flaass). We present some general results connecting the
structure of a coloring group to the structure of its graph coloring, providing
graph-theoretic characterizations of the centralizer and primitivity of a
coloring group. We apply these results particularly to generalized toggle
groups arising from trees as well as coloring groups arising from the
independence posets introduced by Thomas and Williams.
| arxiv topic:math.CO math.GR |
arxiv_dataset-196502312.03192 | Modeling Structure and Country-specific Heterogeneity in
Misclassification Matrices of Verbal Autopsy-based Cause of Death Classifiers
stat.ME stat.AP
Verbal autopsy (VA) algorithms are routinely used to determine
individual-level causes of death (COD) in many low-and-middle-income countries,
which are then aggregated to derive population-level cause-specific mortality
fractions (CSMF), essential to informing public health policies. However, VA
algorithms frequently misclassify COD and introduce bias in CSMF estimates. A
recent method, VA-calibration, can correct for this bias using a VA
misclassification matrix estimated from paired data on COD from both VA and
minimally invasive tissue sampling (MITS) from the Child Health and Mortality
Prevention Surveillance (CHAMPS) Network. Due to the limited sample size,
CHAMPS data are pooled across all countries, implicitly assuming that the
misclassification rates are homogeneous.
In this research, we show that the VA misclassification matrices are
substantially heterogeneous across countries, thereby biasing the
VA-calibration. We develop a coherent framework for modeling country-specific
misclassification matrices in data-scarce settings. We first introduce a novel
base model based on two latent mechanisms: intrinsic accuracy and systematic
preference to parsimoniously characterize misclassifications. We prove that
they are identifiable from the data and manifest as a form of invariance in
certain misclassification odds, a pattern evident in the CHAMPS data. Then we
expand from this base model, adding higher complexity and country-specific
heterogeneity via interpretable effect sizes. Shrinkage priors balance the
bias-variance tradeoff by adaptively favoring simpler models. We publish
uncertainty-quantified estimates of VA misclassification rates for 6 countries.
This effort broadens VA-calibration's future applicability and strengthens
ongoing efforts of using VA for mortality surveillance.
| arxiv topic:stat.ME stat.AP |
arxiv_dataset-196512312.03292 | Enhancing Molecular Property Prediction via Mixture of Collaborative
Experts
cs.LG cs.MA q-bio.QM
Molecular Property Prediction (MPP) task involves predicting biochemical
properties based on molecular features, such as molecular graph structures,
contributing to the discovery of lead compounds in drug development. To address
data scarcity and imbalance in MPP, some studies have adopted Graph Neural
Networks (GNN) as an encoder to extract commonalities from molecular graphs.
However, these approaches often use a separate predictor for each task,
neglecting the shared characteristics among predictors corresponding to
different tasks. In response to this limitation, we introduce the GNN-MoCE
architecture. It employs the Mixture of Collaborative Experts (MoCE) as
predictors, exploiting task commonalities while confronting the homogeneity
issue in the expert pool and the decision dominance dilemma within the expert
group. To enhance expert diversity for collaboration among all experts, the
Expert-Specific Projection method is proposed to assign a unique projection
perspective to each expert. To balance decision-making influence for
collaboration within the expert group, the Expert-Specific Loss is presented to
integrate individual expert loss into the weighted decision loss of the group
for more equitable training. Benefiting from the enhancements of MoCE in expert
creation, dynamic expert group formation, and experts' collaboration, our model
demonstrates superior performance over traditional methods on 24 MPP datasets,
especially in tasks with limited data or high imbalance.
| arxiv topic:cs.LG cs.MA q-bio.QM |
arxiv_dataset-196522312.03392 | O'Neill's Theorem for Games
cs.GT
We present an analog of O'Neill's Theorem (Theorem 5.2 in [17]) for finite
games, which reveals some of the structure of equilibria under payoff
perturbations in finite games.
| arxiv topic:cs.GT |
arxiv_dataset-196532312.03492 | Learning From Scenarios for Stochastic Repairable Scheduling
cs.LG cs.AI
When optimizing problems with uncertain parameter values in a linear
objective, decision-focused learning enables end-to-end learning of these
values. We are interested in a stochastic scheduling problem, in which
processing times are uncertain, which brings uncertain values in the
constraints, and thus repair of an initial schedule may be needed. Historical
realizations of the stochastic processing times are available. We show how
existing decision-focused learning techniques based on stochastic smoothing can
be adapted to this scheduling problem. We include an extensive experimental
evaluation to investigate in which situations decision-focused learning
outperforms the state of the art for such situations: scenario-based stochastic
optimization.
| arxiv topic:cs.LG cs.AI |
arxiv_dataset-196542312.03592 | Analytical solutions for quantum radiation reaction in high-intensity
lasers
physics.plasm-ph hep-ph
While the Landau-Lifshitz equation, which describes classical radiation
reaction, can be solved exactly and analytically for a charged particle
accelerated by a plane electromagnetic wave, no such solutions are available
for quantum radiation reaction (the recoil arising from the successive,
incoherent emission of hard photons). Yet upcoming experiments with
ultrarelativistic electron beams and high-intensity lasers will explore the
regime where both radiation-reaction and quantum effects are important. Here we
present analytical solutions for the mean and variance of the energy
distribution of an electron beam that collides with a pulsed plane
electromagnetic wave, which are obtained by means of a perturbative expansion
in the quantum parameter $\chi_0$. These solutions capture both the quantum
reduction in the radiated power and stochastic broadening, and are shown to be
accurate across the range of experimentally relevant collision parameters, i.e.
GeV-class electron beams and laser amplitudes $a_0 \lesssim 200$.
| arxiv topic:physics.plasm-ph hep-ph |
arxiv_dataset-196552312.03692 | Memory Triggers: Unveiling Memorization in Text-To-Image Generative
Models through Word-Level Duplication
cs.CR cs.CV cs.LG
Diffusion-based models, such as the Stable Diffusion model, have
revolutionized text-to-image synthesis with their ability to produce
high-quality, high-resolution images. These advancements have prompted
significant progress in image generation and editing tasks. However, these
models also raise concerns due to their tendency to memorize and potentially
replicate exact training samples, posing privacy risks and enabling adversarial
attacks. Duplication in training datasets is recognized as a major factor
contributing to memorization, and various forms of memorization have been
studied so far. This paper focuses on two distinct and underexplored types of
duplication that lead to replication during inference in diffusion-based
models, particularly in the Stable Diffusion model. We delve into these
lesser-studied duplication phenomena and their implications through two case
studies, aiming to contribute to the safer and more responsible use of
generative models in various applications.
| arxiv topic:cs.CR cs.CV cs.LG |
arxiv_dataset-196562312.03792 | PCDP-SGD: Improving the Convergence of Differentially Private SGD via
Projection in Advance
cs.CR cs.LG
The paradigm of Differentially Private SGD~(DP-SGD) can provide a theoretical
guarantee for training data in both centralized and federated settings.
However, the utility degradation caused by DP-SGD limits its wide application
in high-stakes tasks, such as medical image diagnosis. In addition to the
necessary perturbation, the convergence issue is attributed to the information
loss on the gradient clipping. In this work, we propose a general framework
PCDP-SGD, which aims to compress redundant gradient norms and preserve more
crucial top gradient components via projection operation before gradient
clipping. Additionally, we extend PCDP-SGD as a fundamental component in
differential privacy federated learning~(DPFL) for mitigating the data
heterogeneous challenge and achieving efficient communication. We prove that
pre-projection enhances the convergence of DP-SGD by reducing the dependence of
clipping error and bias to a fraction of the top gradient eigenspace, and in
theory, limits cross-client variance to improve the convergence under
heterogeneous federation. Experimental results demonstrate that PCDP-SGD
achieves higher accuracy compared with state-of-the-art DP-SGD variants in
computer vision tasks. Moreover, PCDP-SGD outperforms current federated
learning frameworks when DP is guaranteed on local training sets.
| arxiv topic:cs.CR cs.LG |
arxiv_dataset-196572312.03892 | A minimal model for the role of the reaction rate on the initiation and
self-sustenance of curved detonations
physics.flu-dyn
A minimal model for curved detonations is studied, illustrating the role of
the reaction rate on the detonation speed and its propagation limits. The model
is based on a simple extension of the minimal Fickett toy model for detonations
based on the kinematic wave equation. The use of a simple depletion rate
conditioned on the shock speed serves to illustrate its role in the
quasi-steady structure of curved waves and their initiation from a strong blast
wave. Calculations of strong initiation from a self-similar explosion
illustrate the various asymptotic regimes of the transition to self-sustenance
and their link to the steady wave structure. We recover the asymptotic regimes
of detonation formation suggested by He and Clavin and modelled in the context
of Detonation Shock Dynamics by Stewart and collaborators. Following an
analysis using the shock change equation, we identify a unique criterion that
permits to infer the critical energy for initiation from the competition
between energy release and geometric decay.
| arxiv topic:physics.flu-dyn |
arxiv_dataset-196582312.03992 | Analytic Approach to the Non-Preemptive Markovian Priority Queue
math.PR
Explicit and exact results are obtained for the joint queue-length
distribution for the two-level non-preemptive Markovian priority queue.
Marginal distributions are derived for the general multi-level problem. The
results are based on a representation of the joint queue-length probability
mass function as a single-variable complex contour integral, that reduces to a
real integral on a finite interval arising from a cut on the real axis. Both
numerical quadrature rules and exact finite sums, involving Legendre
polynomials and their generalization, are presented for the joint and marginal
distributions. A high level of accuracy is demonstrated across the entire
ergodic region. Relationships are established with the waiting-time
distributions. Asymptotic behaviour in the large queue-length regime is
extracted.
| arxiv topic:math.PR |
arxiv_dataset-196592312.04092 | Data stewardship: case studies from North-American, Dutch, and Finnish
universities
cs.DL
Purpose - As national legislation, federated national services, institutional
policies and institutional research service arrangements may differ, data
stewardship programs may be organized differently in higher education
institutions across the world. This work seeks to elaborate the picture of
different data stewardship programs running in different institutional and
national research environments. Design/methodology/approach - Utilizing a case
study design, this study described three distinct data stewardship programs
from Purdue University (United States), Delft Technical University
(Netherlands) and Aalto University (Finland). In addition, this work
investigated the institutional and national research environments of the
programs. The focus was on initiatives led by academic libraries or similar
services.Findings - This work demonstrates that data stewardship programs may
be organized differently within varying national and institutional contexts.
The data stewardship programs varied in terms of roles, organization and
funding structures. Furthermore, policies and legislation, organizational
structures, and national infrastructures differed. Originality - This work
broadens the current literature on data stewardship by not only providing
detailed descriptions of three distinct data stewardship programs, but also
highlighting how research environments may affect their organization. We
present a summary of key factors in the organization of data stewardship
programs. Research limitations/implications - The data stewardship programs and
their contexts develop, and the descriptions presented in this work should be
considered as snapshots.
| arxiv topic:cs.DL |
arxiv_dataset-196602312.04192 | Convergence Rate Analysis of Continuous- and Discrete-Time Smoothing
Gradient Algorithms
math.OC
This paper addresses the gradient flow -- the continuous-time representation
of the gradient method -- with the smooth approximation of a non-differentiable
objective function and presents convergence analysis framework. Similar to the
gradient method, the gradient flow is inapplicable to the non-differentiable
function minimization; therefore, this paper addresses the smoothing gradient
method, which exploits a decreasing smoothing parameter sequence in the smooth
approximation. The convergence analysis is presented using conventional
Lyapunov-function-based techniques, and a Lyapunov function applicable to both
strongly convex and non-strongly convex objective functions is provided by
taking into consideration the effect of the smooth approximation. Based on the
equivalence of the stepsize in the smoothing gradient method and the
discretization step in the forward Euler scheme for the numerical integration
of the smoothing gradient flow, the sample values of the exact solution of the
smoothing gradient flow are compared with the state variable of the smoothing
gradient method, and the equivalence of the convergence rates is shown.
| arxiv topic:math.OC |
arxiv_dataset-196612312.04292 | Origin of slow-drift shadow bursts in Jovian decameter radio emission
with quasi-harmonic structure
astro-ph.HE astro-ph.EP physics.space-ph
An explanation is proposed for the appearance of slowly drift shadow bursts
in the dynamic spectrum of Jupiter against the background of decameter radio
emission with a quasi-harmonic structure. Background radio emission is caused
by hot ions with a loss cone type distribution function, which generate ion
cyclotron waves due to the effect of double plasma resonance. A flow of hot
ions with a distribution function of the Maxwell type is injected into the
source region, fills the loss cone of generating ions and interrupts the
generation of ion cyclotron waves due to the filling of the loss cone. The
condition under which instability breaks down is obtained, and the optimal
values of the parameters of the injected ions necessary for the occurrence of
bursts in absorption are determined.
| arxiv topic:astro-ph.HE astro-ph.EP physics.space-ph |
arxiv_dataset-196622312.04392 | Contextual Subspace Variational Quantum Eigensolver Calculation of the
Dissociation Curve of Molecular Nitrogen on a Superconducting Quantum
Computer
quant-ph
In this work we present an experimental demonstration of the Contextual
Subspace Variational Quantum Eigensolver on superconducting quantum hardware.
In particular, we compute the potential energy curve for molecular nitrogen,
where a dominance of static correlation in the dissociation limit proves
challenging for many conventional quantum chemistry techniques. Our quantum
simulations retain good agreement with the full configuration interaction
energy in the chosen STO-3G basis, outperforming all benchmarked
single-reference wavefunction techniques in capturing the bond-breaking
appropriately. Moreover, our methodology is competitive with several
multiconfigurational approaches, but at a considerable saving of quantum
resource, meaning larger active spaces can be treated for a fixed qubit
allowance. To achieve this result we deploy an error mitigation/suppression
strategy comprised of dynamical decoupling, measurement-error mitigation and
zero-noise extrapolation, in addition to circuit parallelization that not only
provides passive averaging of noise but improves the effective shot-yield to
reduce the measurement overhead. Furthermore, we introduce a modification to
previous adaptive ansatz construction algorithms that incorporates
hardware-awareness into our variational circuits to minimize the transpilation
cost for the target qubit topology.
| arxiv topic:quant-ph |
arxiv_dataset-196632312.04492 | Ergodic theorems for continuous-time quantum walks on crystal lattices
and the torus
math-ph math.MP math.SP
We give several quantum dynamical analogs of the classical Kronecker-Weyl
theorem, which says that the trajectory of free motion on the torus along
almost every direction tends to equidistribute. As a quantum analog, we study
the quantum walk $\exp(-i t \Delta) \psi$ starting from a localized initial
state $\psi$. Then the flow will be ergodic if this evolved state becomes
equidistributed as time goes on. We prove that this is indeed the case for
evolutions on the flat torus, provided we start from a point mass, and we prove
discrete analogs of this result for crystal lattices. On some periodic graphs,
the mass spreads out non-uniformly, on others it stays localized. Finally, we
give examples of quantum evolutions on the sphere which do not equidistribute.
| arxiv topic:math-ph math.MP math.SP |
arxiv_dataset-196642312.04592 | Current induced magnetisation in metal without space-inversion symmetry
cond-mat.str-el cond-mat.mes-hall
Magneto-electric effect, that is an appearance of magnetisation induced by
electric current is allowed by symmetry in metals with crystal structure
without space inversion. The microscopic origin of this effect is spin-orbit
coupling of electrons with a non-centrosymmetric crystal lattice lifting spin
degeneracy of electron energy and mixing spin and orbital degrees of freedom.
The presented calculation of magnetisation induced by current based on the
application of kinetic equation for the matrix distribution function of
electrons occupying the states in two bands split by the spin-orbit
interaction.
| arxiv topic:cond-mat.str-el cond-mat.mes-hall |
arxiv_dataset-196652312.04692 | Diffence: Fencing Membership Privacy With Diffusion Models
cs.CR cs.CV cs.LG
Deep learning models, while achieving remarkable performances, are vulnerable
to membership inference attacks (MIAs). Although various defenses have been
proposed, there is still substantial room for improvement in the
privacy-utility trade-off. In this work, we introduce a novel defense framework
against MIAs by leveraging generative models. The key intuition of our defense
is to remove the differences between member and non-member inputs, which is
exploited by MIAs, by re-generating input samples before feeding them to the
target model. Therefore, our defense, called DIFFENCE, works pre inference,
which is unlike prior defenses that are either training-time or post-inference
time.
A unique feature of DIFFENCE is that it works on input samples only, without
modifying the training or inference phase of the target model. Therefore, it
can be cascaded with other defense mechanisms as we demonstrate through
experiments. DIFFENCE is designed to preserve the model's prediction labels for
each sample, thereby not affecting accuracy. Furthermore, we have empirically
demonstrated it does not reduce the usefulness of confidence vectors. Through
extensive experimentation, we show that DIFFENCE can serve as a robust
plug-n-play defense mechanism, enhancing membership privacy without
compromising model utility. For instance, DIFFENCE reduces MIA accuracy against
an undefended model by 15.8\% and attack AUC by 14.0\% on average across three
datasets, all without impacting model utility. By integrating DIFFENCE with
prior defenses, we can achieve new state-of-the-art performances in the
privacy-utility trade-off. For example, when combined with the state-of-the-art
SELENA defense it reduces attack accuracy by 9.3\%, and attack AUC by 10.0\%.
DIFFENCE achieves this by imposing a negligible computation overhead, adding
only 57ms to the inference time per sample processed on average.
| arxiv topic:cs.CR cs.CV cs.LG |
arxiv_dataset-196662312.04792 | Playing Large Games with Oracles and AI Debate
cs.GT cs.AI
We consider regret minimization in repeated games with a very large number of
actions. Such games are inherent in the setting of AI Safety via Debate
\cite{irving2018ai}, and more generally games whose actions are language-based.
Existing algorithms for online game playing require per-iteration computation
polynomial in the number of actions, which can be prohibitive for large games.
We thus consider oracle-based algorithms, as oracles naturally model access
to AI agents. With oracle access, we characterize when internal and external
regret can be minimized efficiently. We give a novel efficient algorithm for
simultaneous external and internal regret minimization whose regret depends
logarithmically on the number of actions. We conclude with experiments in the
setting of AI Safety via Debate that shows the benefit of insights from our
algorithmic analysis.
| arxiv topic:cs.GT cs.AI |
arxiv_dataset-196672312.04892 | Floquet engineering of many-body states by the ponderomotive potential
cond-mat.str-el cond-mat.mes-hall cond-mat.other cond-mat.supr-con quant-ph
The ponderomotive force is an effective static force that a particle feels in
an oscillating field, whose static potential may be called the ponderomotive
potential. We generalize this notion to periodically driven quantum many-body
systems, and propose it as a convenient tool to engineer their non-equilibrium
steady states beyond the single particle level. Applied to materials driven by
light, the ponderomotive potential is intimately related to the equilibrium
optical conductivity, which is enhanced close to resonances. We show that the
ponderomotive potential from the incident light may be used to induce exciton
condensates in semiconductors, to generate attractive interactions leading to
superconductivity in certain electron-phonon systems, and to create additional
free energy minima in systems with charge/spin/excitonic orders. These effects
are presented with experimentally relevant parameters.
| arxiv topic:cond-mat.str-el cond-mat.mes-hall cond-mat.other cond-mat.supr-con quant-ph |
arxiv_dataset-196682312.04992 | PFLlib: A Beginner-Friendly and Comprehensive Personalized Federated
Learning Library and Benchmark
cs.LG cs.DC
Amid the ongoing advancements in Federated Learning (FL), a machine learning
paradigm that allows collaborative learning with data privacy protection,
personalized FL (pFL)has gained significant prominence as a research direction
within the FL domain. Whereas traditional FL (tFL) focuses on jointly learning
a global model, pFL aims to balance each client's global and personalized goals
in FL settings. To foster the pFL research community, we started and built
PFLlib, a comprehensive pFL library with an integrated benchmark platform. In
PFLlib, we implemented 37 state-of-the-art FL algorithms (8 tFL algorithms and
29 pFL algorithms) and provided various evaluation environments with three
statistically heterogeneous scenarios and 24 datasets. At present, PFLlib has
gained more than 1600 stars and 300 forks on GitHub.
| arxiv topic:cs.LG cs.DC |
arxiv_dataset-196692312.05092 | INSPECT: Intrinsic and Systematic Probing Evaluation for Code
Transformers
cs.SE cs.LG
Pre-trained models of source code have recently been successfully applied to
a wide variety of Software Engineering tasks; they have also seen some
practical adoption in practice, e.g. for code completion. Yet, we still know
very little about what these pre-trained models learn about source code. In
this article, we use probing--simple diagnostic tasks that do not further train
the models--to discover to what extent pre-trained models learn about specific
aspects of source code. We use an extensible framework to define 15 probing
tasks that exercise surface, syntactic, structural and semantic characteristics
of source code. We probe 8 pre-trained source code models, as well as a natural
language model (BERT) as our baseline. We find that models that incorporate
some structural information (such as GraphCodeBERT) have a better
representation of source code characteristics. Surprisingly, we find that for
some probing tasks, BERT is competitive with the source code models, indicating
that there are ample opportunities to improve source-code specific pre-training
on the respective code characteristics. We encourage other researchers to
evaluate their models with our probing task suite, so that they may peer into
the hidden layers of the models and identify what intrinsic code
characteristics are encoded.
| arxiv topic:cs.SE cs.LG |
arxiv_dataset-196702312.05192 | Force Matching and Iterative Boltzmann Inversion Coarse Grained Force
Fields for ZIF-8
cond-mat.mtrl-sci
Despite the intense activity at the electronic and atomistic resolutions,
coarse grained (CG) modeling of MOFs remains largely unexplored. One of the
main reasons for this is the lack of adequate CG force fields. In this work, we
present Iterative Boltzmann Inversion (IBI) and Force Matching (FM) force
fields for modeling ZIF-8 in three different coarse grained resolutions. Their
ability of reproducing structure, elastic tensor and thermal expansion is
evaluated and compared with that of MARTINI force-fields considered in previous
work.[C. M. S. Alvares et al, J. Chem. Phys., 158, 194107 (2023).] Moreover,
MARTINI and FM are evaluated in their ability of depicting the swing effect, a
subtle phase transition ZIF-8 undergoes when loaded with guest molecules.
Overall, we found that all our force fields reproduce structure reasonably
well. Elastic constants and volume expansion results are analyzed and the
technical and conceptual challenges in reproducing them are explained. Force
matching exhibits promising results for capturing the swing effect. This is the
first time these CG methods, widely applied in polymer and biomolecules
communities, are deployed to model porous solids. We highlight the challenges
of fitting CG force fields for these materials. This work opens the door to a
whole new line of developments in the field of modeling MOFs and other porous
crystalline solids.
| arxiv topic:cond-mat.mtrl-sci |
arxiv_dataset-196712312.05292 | Discovery of a large and faint nebula at the Triangulum galaxy
astro-ph.GA
We report the discovery of a previously uncatalogued arch-shaped filamentary
nebula at the outer part of the Triangulum galaxy (M33) centred at R.A. =
1h34m25s, Dec = +30d20m17s (ICRS). This discovery stems from meticulous
observations employing deep exposures of M33, using both H-alpha and [OIII]
narrow-band filters. The nebula, designated as "Roig1 Prades Sky", exhibits an
H-alpha surface brightness of 23.9 mag/arcsec2. Its sky projected location is
21 arcmin away from the M33 galactic centre towards the southeast direction
with an extent of 120 by 440 pc. Deep spectroscopic observations are required
to unveil its real nature.
| arxiv topic:astro-ph.GA |
arxiv_dataset-196722312.05392 | The logic of NTQR evaluations of noisy AI agents: Complete postulates
and logically consistent error correlations
cs.AI
In his "ship of state" allegory (\textit{Republic}, Book VI, 488) Plato poses
a question -- how can a crew of sailors presumed to know little about the art
of navigation recognize the true pilot among them? The allegory argues that a
simple majority voting procedure cannot safely determine who is most qualified
to pilot a ship when the voting members are ignorant or biased. We formalize
Plato's concerns by considering the problem in AI safety of monitoring noisy AI
agents in unsupervised settings. An algorithm evaluating AI agents using
unlabeled data would be subject to the evaluation dilemma - how would we know
the evaluation algorithm was correct itself? This endless validation chain can
be avoided by considering purely algebraic functions of the observed responses.
We can construct complete postulates than can prove or disprove the logical
consistency of any grading algorithm. A complete set of postulates exists
whenever we are evaluating $N$ experts that took $T$ tests with $Q$ questions
with $R$ responses each. We discuss evaluating binary classifiers that have
taken a single test - the $(N,T=1,Q,R=2)$ tests. We show how some of the
postulates have been previously identified in the ML literature but not
recognized as such - the \textbf{agreement equations} of Platanios. The
complete postulates for pair correlated binary classifiers are considered and
we show how it allows for error correlations to be quickly calculated. An
algebraic evaluator based on the assumption that the ensemble is error
independent is compared with grading by majority voting on evaluations using
the \uciadult and and \texttt{two-norm} datasets. Throughout, we demonstrate
how the formalism of logical consistency via algebraic postulates of evaluation
can help increase the safety of machines using AI algorithms.
| arxiv topic:cs.AI |
arxiv_dataset-196732312.05492 | cuSZ-$i$: High-Ratio Scientific Lossy Compression on GPUs with Optimized
Multi-Level Interpolation
cs.DC
Error-bounded lossy compression is a critical technique for significantly
reducing scientific data volumes. Compared to CPU-based compressors, GPU-based
compressors exhibit substantially higher throughputs, fitting better for
today's HPC applications. However, the critical limitations of existing
GPU-based compressors are their low compression ratios and qualities, severely
restricting their applicability. To overcome these, we introduce a new
GPU-based error-bounded scientific lossy compressor named cuSZ-$i$, with the
following contributions: (1) A novel GPU-optimized interpolation-based
prediction method significantly improves the compression ratio and
decompression data quality. (2) The Huffman encoding module in cuSZ-$i$ is
optimized for better efficiency. (3) cuSZ-$i$ is the first to integrate the
NVIDIA Bitcomp-lossless as an additional compression-ratio-enhancing module.
Evaluations show that cuSZ-$i$ significantly outperforms other latest GPU-based
lossy compressors in compression ratio under the same error bound (hence, the
desired quality), showcasing a 476% advantage over the second-best. This leads
to cuSZ-$i$'s optimized performance in several real-world use cases.
| arxiv topic:cs.DC |
arxiv_dataset-196742312.05592 | Improving reconstructions in nanotomography for homogeneous materials
via mathematical optimization
cond-mat.mtrl-sci math.OC
Compressed sensing is an image reconstruction technique to achieve
high-quality results from limited amount of data. In order to achieve this, it
utilizes prior knowledge about the samples that shall be reconstructed.
Focusing on image reconstruction in nanotomography, this work proposes
enhancements by including additional problem-specific knowledge. In more
detail, we propose further classes of algebraic inequalities that are added to
the compressed sensing model. The first consists in a valid upper bound on the
pixel brightness. It only exploits general information about the projections
and is thus applicable to a broad range of reconstruction problems. The second
class is applicable whenever the sample material is of roughly homogeneous
composition. The model favors a constant density and penalizes deviations from
it. The resulting mathematical optimization models are algorithmically
tractable and can be solved to global optimality by state-of-the-art available
implementations of interior point methods. In order to evaluate the novel
models, obtained results are compared to existing image reconstruction methods,
tested on simulated and experimental data sets. The experimental data comprise
one 360{\deg} electron tomography tilt series of a macroporous zeolite particle
and one absorption contrast nano X-ray computed tomography (nano-CT) data set
of a copper microlattice structure. The enriched models are optimized quickly
and show improved reconstruction quality, outperforming the existing models.
Promisingly, our approach yields superior reconstruction results, particularly
when information about the samples is available for a small number of tilt
angles only
| arxiv topic:cond-mat.mtrl-sci math.OC |
arxiv_dataset-196752312.05692 | Decay estimates for Cayley transforms and inverses of semigroup
generators via the $\mathcal{B}$-calculus
math.FA cs.NA math.NA
Let $-A$ be the generator of a bounded $C_0$-semigroup $(e^{-tA})_{t \geq 0}$
on a Hilbert space. First we study the long-time asymptotic behavior of the
Cayley transform $V_{\omega}(A) := (A-\omega I) (A+\omega I)^{-1}$ with $\omega
>0$. We give a decay estimate for $\|V_{\omega}(A)^nA^{-1}\|$ when
$(e^{-tA})_{t \geq 0}$ is polynomially stable. Considering the case where the
parameter $\omega$ varies, we estimate $\|(\prod_{k=1}^n
V_{\omega_k}(A))A^{-1}\|$ for exponentially stable $C_0$-semigroups
$(e^{-tA})_{t \geq 0}$. Next we show that if the generator $-A$ of the bounded
$C_0$-semigroup has a bounded inverse, then $\sup_{t \geq 0} \|e^{-tA^{-1}}
A^{-\alpha} \| < \infty$ for all $\alpha >0$. We also present an estimate for
the rate of decay of $\|e^{-tA^{-1}} A^{-1} \|$, assuming that $(e^{-tA})_{t
\geq 0}$ is polynomially stable. To obtain these results, we use operator norm
estimates offered by a functional calculus called the $\mathcal{B}$-calculus.
| arxiv topic:math.FA cs.NA math.NA |
arxiv_dataset-196762312.05792 | Take an Irregular Route: Enhance the Decoder of Time-Series Forecasting
Transformer
cs.LG cs.AI
With the development of Internet of Things (IoT) systems, precise long-term
forecasting method is requisite for decision makers to evaluate current
statuses and formulate future policies. Currently, Transformer and MLP are two
paradigms for deep time-series forecasting and the former one is more
prevailing in virtue of its exquisite attention mechanism and encoder-decoder
architecture. However, data scientists seem to be more willing to dive into the
research of encoder, leaving decoder unconcerned. Some researchers even adopt
linear projections in lieu of the decoder to reduce the complexity. We argue
that both extracting the features of input sequence and seeking the relations
of input and prediction sequence, which are respective functions of encoder and
decoder, are of paramount significance. Motivated from the success of FPN in CV
field, we propose FPPformer to utilize bottom-up and top-down architectures
respectively in encoder and decoder to build the full and rational hierarchy.
The cutting-edge patch-wise attention is exploited and further developed with
the combination, whose format is also different in encoder and decoder, of
revamped element-wise attention in this work. Extensive experiments with six
state-of-the-art baselines on twelve benchmarks verify the promising
performances of FPPformer and the importance of elaborately devising decoder in
time-series forecasting Transformer. The source code is released in
https://github.com/OrigamiSL/FPPformer.
| arxiv topic:cs.LG cs.AI |
arxiv_dataset-196772312.05892 | Quasiparticle dynamics in a superconducting qubit irradiated by a
localized infrared source
quant-ph cond-mat.supr-con
A known source of decoherence in superconducting qubits is the presence of
broken Cooper pairs, or quasiparticles. These can be generated by high-energy
radiation, either present in the environment or purposefully introduced, as in
the case of some hybrid quantum devices. Here, we systematically study the
properties of a transmon qubit under illumination by focused infrared radiation
with various powers, durations, and spatial locations. Despite the high energy
of incident photons, our observations agree well with a model of low-energy
quasiparticle dynamics dominated by trapping. This technique can be used for
understanding and potentially mitigating the effects of high-energy radiation
on superconducting circuits with a variety of geometries and materials.
| arxiv topic:quant-ph cond-mat.supr-con |
arxiv_dataset-196782312.05992 | Probing the Interactions of Axion-Like Particles with Electroweak Bosons
and the Higgs Boson in the High Energy Regime at LHC
hep-ph
We study the interactions of Axion-Like Particles (ALPs) with the Standard
Model particles, aiming to probe their phenomenology via non-resonant searches
at the LHC. These interactions are mediated by higher dimensional effective
operators within two possible frameworks of linearly and non-linearly realised
electroweak symmetry breaking. We consider the ALPs to be light enough to be
produced on-shell and exploit their derivative couplings with the SM Higgs
boson and the gauge bosons. We will use the high momentum transfer processes,
namely $hZ, Z\gamma, WW$ and $WW\gamma$ production from $pp$ collisions. We
derive upper limits on the gauge-invariant interactions of ALPs with the
electroweak bosons and/or Higgs boson that contribute to these processes, from
the re-interpretation of the latest Run 2 available LHC data. The constraints
we obtain are strong for ALP masses below 100 GeV. These allowed effective
interactions in the ALP parameter space yield better significance at HL-LHC and
thus, offer promising avenues for subsequent studies. Furthermore, we augment
our cut-based analysis with gradient-boosted decision trees, which improve the
statistical significance distinctly across these interaction channels. We
briefly compare the results with the complementary probe of these couplings via
direct production of ALPs in association with the Higgs boson or a vector
boson.
| arxiv topic:hep-ph |
arxiv_dataset-196792312.06092 | Analysis of Synchrosqueezed Transforms and Application Perspectives
eess.SP
High-resolution time-frequency (TF) analysis plays crucial role in
characterizing multicomponent signal (MCSs) and estimating oscillatory
properties. Linear time-frequency representations (TFRs) such as classical
short-time Fourier transform (STFT) and continuous wavelet transform (CWT)
incur constrained TF resolution and energy diffusion in both time and frequency
direction. The synchrosqueezing transform (SST) represents a powerful sparse
reassignment method that allows component reconstruction. This work introduces
SST as extension to STFT and CWT and illustrates corresponding advantages of
sharpening TFRs and recovery of instantaneous components. The SST effectiveness
is assessed in practical situations that involve comparing STFT-based and
CWT-based versions of synthetic data and also applying SST to optimize deep
learning (DL) prediction model. It is demonstrated how SST achieves promising
results in terms of improving TFR readability and increasing accuracy of
DL-based prediction models.
| arxiv topic:eess.SP |
arxiv_dataset-196802312.06192 | NutritionVerse-Synth: An Open Access Synthetically Generated 2D Food
Scene Dataset for Dietary Intake Estimation
cs.CV
Manually tracking nutritional intake via food diaries is error-prone and
burdensome. Automated computer vision techniques show promise for dietary
monitoring but require large and diverse food image datasets. To address this
need, we introduce NutritionVerse-Synth (NV-Synth), a large-scale synthetic
food image dataset. NV-Synth contains 84,984 photorealistic meal images
rendered from 7,082 dynamically plated 3D scenes. Each scene is captured from
12 viewpoints and includes perfect ground truth annotations such as RGB, depth,
semantic, instance, and amodal segmentation masks, bounding boxes, and detailed
nutritional information per food item. We demonstrate the diversity of NV-Synth
across foods, compositions, viewpoints, and lighting. As the largest
open-source synthetic food dataset, NV-Synth highlights the value of
physics-based simulations for enabling scalable and controllable generation of
diverse photorealistic meal images to overcome data limitations and drive
advancements in automated dietary assessment using computer vision. In addition
to the dataset, the source code for our data generation framework is also made
publicly available at https://saeejithnair.github.io/nvsynth.
| arxiv topic:cs.CV |
arxiv_dataset-196812312.06292 | HoLLiE C -- A Multifunctional Bimanual Mobile Robot Supporting Versatile
Care Applications
cs.RO
Care robotics as a research field has developed a lot in recent years, driven
by the rapidly increasing need for it. However, these technologies are mostly
limited to a very concrete and usually relatively simple use case. The bimanual
robot House of Living Labs intelligent Escort (HoLLiE) includes an
omnidirectional mobile platform. This paper presents how HoLLiE is adapted, by
flexible software and hardware modules, for different care applications. The
design goal of HoLLiE was to be human-like but abstract enough to ensure a high
level of acceptance, which is very advantageous for its use in hospitals. After
a short retrospect of previous generations of HoLLiE, it is highlighted how the
current version is equipped with a variety of additional sensors and actuators
to allow a wide range of possible applications. Then, the software stack of
HoLLiE is depicted, with the focus on navigation and force sensitive intention
recognition.
| arxiv topic:cs.RO |
arxiv_dataset-196822312.06392 | Feeding and feedback processes in the Spiderweb proto-intracluster
medium
astro-ph.GA astro-ph.CO
We present the detailed analysis of the thermal, diffuse emission of the
proto-intracluster medium (ICM) detected in the halo of the Spiderweb Galaxy at
z=2.16, within a radius of $\sim$ 150 kpc. We combined deep X-ray data from
Chandra and millimeter observations of the Sunyaev-Zeldovich (SZ) effect
obtained by ALMA. Thanks to independent measurements of the pressure profile
from ALMA SZ observation and the electron density profile from the available
X-ray data, we derived, for the first time, the temperature profile in the ICM
of a z>2 protocluster. It reveals the presence of a strong cool core
(comparable to the local ones) that may host a significant mass deposition
flow, consistent with measured local star formation values. We also find mild
evidence of an asymmetry in the X-ray surface brightness distribution, which
may be tentatively associated with a cavity carved into the proto-ICM by the
radio jets or, alternatively, may be due to the young dynamical status of the
halo. The cooling time of baryons in the core of the Spiderweb Protocluster is
estimated to be $\sim$ 0.1 Gyr, implying that the baryon cycle in the first
stages of the protocluster formation is characterised by a high-duty cycle and
a very active environment. In the case of the Spiderweb protocluster, we are
witnessing the presence of a strongly peaked core that is possibily hosting a
cooling flow with a mass deposition rate up to 250-1000 $M_{\odot}$/yr,
responsible for feeding both the central supermassive black hole and the high
star formation rate observed in the Spiderweb Galaxy. This phase is expected to
be rapidly followed by active galactic nucleus feedback events, whose onset may
have already left an imprint in the radio and X-ray appearance of the Spiderweb
protocluster, eventually driving the ICM into a self-regulated, long-term
evolution in less than one Gyr.
| arxiv topic:astro-ph.GA astro-ph.CO |
arxiv_dataset-196832312.06492 | Phase transitions of LaMnO$_3$ and SrRuO$_3$ from DFT + U based machine
learning force fields simulations
cond-mat.mtrl-sci cond-mat.str-el
Perovskite oxides are known to exhibit many magnetic, electronic and
structural phases as function of doping and temperature. These materials are
theoretically frequently investigated by the DFT+U method, typically in their
ground state structure at $T=0$. We show that by combining machine learning
force fields (MLFFs) and DFT+U based molecular dynamics, it becomes possible to
investigate the crystal structure of complex oxides as function of temperature
and $U$. Here, we apply this method to the magnetic transition metal compounds
LaMnO$_3$ and SrRuO$_3$. We show that the structural phase transition from
orthorhombic to cubic in LaMnO$_3$, which is accompanied by the suppression of
a Jahn-Teller distortion, can be simulated with an appropriate choice of $U$.
For SrRuO$_3$, we show that the sequence of orthorhombic to tetragonal to cubic
crystal phase transitions can be described with great accuracy. We propose that
the $U$ values that correctly capture the temperature-dependent structures of
these complex oxides, can be identified by comparison of the MLFF simulated and
experimentally determined structures.
| arxiv topic:cond-mat.mtrl-sci cond-mat.str-el |
arxiv_dataset-196842312.06592 | Flexible visual prompts for in-context learning in computer vision
cs.CV
In this work, we address in-context learning (ICL) for the task of image
segmentation, introducing a novel approach that adapts a modern Video Object
Segmentation (VOS) technique for visual in-context learning. This adaptation is
inspired by the VOS method's ability to efficiently and flexibly learn objects
from a few examples. Through evaluations across a range of support set sizes
and on diverse segmentation datasets, our method consistently surpasses
existing techniques. Notably, it excels with data containing classes not
encountered during training. Additionally, we propose a technique for support
set selection, which involves choosing the most relevant images to include in
this set. By employing support set selection, the performance increases for all
tested methods without the need for additional training or prompt tuning. The
code can be found at https://github.com/v7labs/XMem_ICL/.
| arxiv topic:cs.CV |
arxiv_dataset-196852312.06692 | Extra Attraction Generated by Spacetime Fluctuations
gr-qc hep-th
We show that, due to the nonlinear nature of gravity, fluctuations in
spacetime curvature generate additional gravitational attraction. This
fluctuation-induced extra attraction was overlooked in the conventional
understanding of the cosmological constant problem. If the quantum vacuum of
matter fields possesses positive energy and negative pressure, it would produce
enormous gravitational repulsion, resulting in a catastrophic explosion of the
universe -- the acceleration of the universe's expansion would exceed the
observed value by some 120 orders of magnitude. We argue that such an enormous
repulsion produced by the violent matter fields vacuum can be completely
suppressed by the even more substantial attraction generated by the zero-point
fluctuations in the spacetime curvature. As a result, the predicted
catastrophic explosion of the universe is averted. Furthermore, at small
microscopic scales, the structure of spacetime becomes locally highly
inhomogeneous and anisotropic. When averaged over large macroscopic scales, the
zero-point fluctuations of spacetime itself could drive the observed slow
acceleration of the universe's expansion through a subtle parametric resonance
effect.
| arxiv topic:gr-qc hep-th |
arxiv_dataset-196862312.06792 | Double points and image of reflection maps
math.AG
A reflection mapping is a singular holomorphic mapping obtained by
restricting the quotient mapping of a complex reflection group. We study the
analytic structure of double point spaces of reflection mappings. In the case
where the image is a hypersurface, we obtain explicit equations for the double
point space and for the image as well. In the case of surfaces in $\C^3$, this
gives a very efficient method to compute the Milnor number and delta invariant
of the double point curve.
| arxiv topic:math.AG |
arxiv_dataset-196872312.06892 | VitalLens: Take A Vital Selfie
cs.CV cs.HC
This report introduces VitalLens, an app that estimates vital signs such as
heart rate and respiration rate from selfie video in real time. VitalLens uses
a computer vision model trained on a diverse dataset of video and physiological
sensor data. We benchmark performance on several diverse datasets, including
VV-Medium, which consists of 289 unique participants. VitalLens outperforms
several existing methods including POS and MTTS-CAN on all datasets while
maintaining a fast inference speed. On VV-Medium, VitalLens achieves mean
absolute errors of 0.71 bpm for heart rate estimation, and 0.76 bpm for
respiratory rate estimation.
| arxiv topic:cs.CV cs.HC |
arxiv_dataset-196882312.06992 | Between the cosmic-ray `knee' and the `ankle': Contribution from star
clusters
astro-ph.HE
We show that massive young star clusters may be possible candidates that can
accelerate Galactic cosmic rays (CRs) in the range of $10^7\hbox{--}10^9$ GeV
(between the `knee' and `ankle'). Various plausible scenarios such as
acceleration at the wind termination shock (WTS), supernova shocks inside these
young star clusters, etc. have been proposed,since it is difficult to
accelerate particles up to the $10^7\hbox{--}10^9$ GeV range in the standard
paradigm of CR acceleration in supernova remnants. We consider a model for the
production of different nuclei in CRs from massive stellar winds using the
observed distribution of young star clusters in the Galactic plane. We present
a detailed calculation of CR transport in the Galaxy, taking into account the
effect of diffusion, interaction losses during propagation, and particle
re-acceleration by old supernova remnants to determine the all-particle CR
spectrum. Using the maximum energy estimate from the Hillas criterion, we argue
that a young massive star cluster can accelerate protons up to a few tens of
PeV. Upon comparison with the observed data, our model requires a CR source
spectrum with an exponential cutoff of $5\times 10^7 Z$ GeV ($50\,Z$~PeV) from
these clusters together with a cosmic-ray injection fraction of $\sim 5\%$ of
the wind kinetic energy. We discuss the possibility of achieving these
requirements in star clusters, and the associated uncertainties, in the context
of considering star clusters as the natural accelerator of the `second
component' of Galactic cosmic rays.
| arxiv topic:astro-ph.HE |
arxiv_dataset-196892312.07092 | Normalized ground states for Schr\"odinger equations on metric graphs
with nonlinear point defects
math.AP
We investigate the existence of normalized ground states for Schr\"odinger
equations on noncompact metric graphs in presence of nonlinear point defects,
described by nonlinear $\delta$-interactions at some of the vertices of the
graph. For graphs with finitely many vertices, we show that ground states exist
for every mass and every $L^2$-subcritical power. For graphs with infinitely
many vertices, we focus on periodic graphs and, in particular, on
$\mathbb{Z}$-periodic graphs and on a prototypical $\mathbb{Z}^2$-periodic
graph, the two-dimensional square grid. We provide a set of results unravelling
nontrivial threshold phenomena both on the mass and on the nonlinearity power,
showing the strong dependence of the ground state problem on the interplay
between the degree of periodicity of the graph, the total number of point
defects and their dislocation in the graph.
| arxiv topic:math.AP |
arxiv_dataset-196902312.07192 | waveSLAM: Empowering Accurate Indoor Mapping Using Off-the-Shelf
Millimeter-wave Self-sensing
cs.NI cs.RO
This paper presents the design, implementation and evaluation of waveSLAM, a
low-cost mobile robot system that uses the millimetre wave (mmWave)
communication devices to enhance the indoor mapping process targeting
environments with reduced visibility or glass/mirror walls. A unique feature of
waveSLAM is that it only leverages existing Commercial-Off-The-Shelf (COTS)
hardware (Lidar and mmWave radios) that are mounted on mobile robots to improve
the accurate indoor mapping achieved with optical sensors. The key intuition
behind the waveSLAM design is that while the mobile robots moves freely, the
mmWave radios can periodically exchange angle and distance estimates between
themselves (self-sensing) by bouncing the signal from the environment, thus
enabling accurate estimates of the target object/material surface. Our
experiments verify that waveSLAM can archive cm-level accuracy with errors
below 22 cm and 20deg in angle orientation which is compatible with Lidar when
building indoor maps.
| arxiv topic:cs.NI cs.RO |
arxiv_dataset-196912312.07292 | Statistically Distinct Plans for Multi-Objective Task Assignment
cs.RO
We study the problem of finding statistically distinct plans for stochastic
planning and task assignment problems such as online multi-robot pickup and
delivery (MRPD) when facing multiple competing objectives. In many real-world
settings robot fleets do not only need to fulfil delivery requests, but also
have to consider auxiliary objectives such as energy efficiency or avoiding
human-centered work spaces. We pose MRPD as a multi-objective optimization
problem where the goal is to find MRPD policies that yield different trade-offs
between given objectives. There are two main challenges: 1) MRPD is
computationally hard, which limits the number of trade-offs that can reasonably
be computed, and 2) due to the random task arrivals, one needs to consider
statistical variance of the objective values in addition to the average. We
present an adaptive sampling algorithm that finds a set of policies which i)
are approximately optimal, ii) approximate the set of all optimal solutions,
and iii) are statistically distinguishable. We prove completeness and adapt a
state-of-the-art MRPD solver to the multi-objective setting for three example
objectives. In a series of simulation experiments we demonstrate the advantages
of the proposed method compared to baseline approaches and show its robustness
in a sensitivity analysis. The approach is general and could be adapted to
other multi-objective task assignment and planning problems under uncertainty.
| arxiv topic:cs.RO |
arxiv_dataset-196922312.07392 | ReRoGCRL: Representation-based Robustness in Goal-Conditioned
Reinforcement Learning
cs.LG cs.AI
While Goal-Conditioned Reinforcement Learning (GCRL) has gained attention,
its algorithmic robustness against adversarial perturbations remains
unexplored. The attacks and robust representation training methods that are
designed for traditional RL become less effective when applied to GCRL. To
address this challenge, we first propose the Semi-Contrastive Representation
attack, a novel approach inspired by the adversarial contrastive attack. Unlike
existing attacks in RL, it only necessitates information from the policy
function and can be seamlessly implemented during deployment. Then, to mitigate
the vulnerability of existing GCRL algorithms, we introduce Adversarial
Representation Tactics, which combines Semi-Contrastive Adversarial
Augmentation with Sensitivity-Aware Regularizer to improve the adversarial
robustness of the underlying RL agent against various types of perturbations.
Extensive experiments validate the superior performance of our attack and
defence methods across multiple state-of-the-art GCRL algorithms. Our tool
ReRoGCRL is available at https://github.com/TrustAI/ReRoGCRL.
| arxiv topic:cs.LG cs.AI |
arxiv_dataset-196932312.07492 | SocialStigmaQA: A Benchmark to Uncover Stigma Amplification in
Generative Language Models
cs.CL cs.AI cs.CY cs.LG
Current datasets for unwanted social bias auditing are limited to studying
protected demographic features such as race and gender. In this work, we
introduce a comprehensive benchmark that is meant to capture the amplification
of social bias, via stigmas, in generative language models. Taking inspiration
from social science research, we start with a documented list of 93 US-centric
stigmas and curate a question-answering (QA) dataset which involves simple
social situations. Our benchmark, SocialStigmaQA, contains roughly 10K prompts,
with a variety of prompt styles, carefully constructed to systematically test
for both social bias and model robustness. We present results for
SocialStigmaQA with two open source generative language models and we find that
the proportion of socially biased output ranges from 45% to 59% across a
variety of decoding strategies and prompting styles. We demonstrate that the
deliberate design of the templates in our benchmark (e.g., adding biasing text
to the prompt or using different verbs that change the answer that indicates
bias) impacts the model tendencies to generate socially biased output.
Additionally, through manual evaluation, we discover problematic patterns in
the generated chain-of-thought output that range from subtle bias to lack of
reasoning.
Warning: This paper contains examples of text which are toxic, biased, and
potentially harmful.
| arxiv topic:cs.CL cs.AI cs.CY cs.LG |
arxiv_dataset-196942312.07592 | Evaluating ChatGPT as a Question Answering System: A Comprehensive
Analysis and Comparison with Existing Models
cs.CL cs.AI
In the current era, a multitude of language models has emerged to cater to
user inquiries. Notably, the GPT-3.5 Turbo language model has gained
substantial attention as the underlying technology for ChatGPT. Leveraging
extensive parameters, this model adeptly responds to a wide range of questions.
However, due to its reliance on internal knowledge, the accuracy of responses
may not be absolute. This article scrutinizes ChatGPT as a Question Answering
System (QAS), comparing its performance to other existing QASs. The primary
focus is on evaluating ChatGPT's proficiency in extracting responses from
provided paragraphs, a core QAS capability. Additionally, performance
comparisons are made in scenarios without a surrounding passage. Multiple
experiments, exploring response hallucination and considering question
complexity, were conducted on ChatGPT. Evaluation employed well-known Question
Answering (QA) datasets, including SQuAD, NewsQA, and PersianQuAD, across
English and Persian languages. Metrics such as F-score, exact match, and
accuracy were employed in the assessment. The study reveals that, while ChatGPT
demonstrates competence as a generative model, it is less effective in question
answering compared to task-specific models. Providing context improves its
performance, and prompt engineering enhances precision, particularly for
questions lacking explicit answers in provided paragraphs. ChatGPT excels at
simpler factual questions compared to "how" and "why" question types. The
evaluation highlights occurrences of hallucinations, where ChatGPT provides
responses to questions without available answers in the provided context.
| arxiv topic:cs.CL cs.AI |
arxiv_dataset-196952312.07692 | Bootstrapping Boundary QED Part I
hep-th
We use the numerical conformal bootstrap to study boundary quantum
electrodynamics, the theory of a four dimensional photon in a half space
coupled to charged conformal matter on the boundary. This system is believed to
be a boundary conformal field theory with an exactly marginal coupling
corresponding to the strength of the interaction between the photon and the
matter degrees of freedom. In part one of this project, we present three
results. We show how the Maxwell equations put severe constraints on boundary
three-point functions involving two currents and a symmetric traceless tensor.
We use semi-definite programming to show that any three dimensional conformal
field theory with a global U(1) symmetry must have a spin two gap less than
about 1.05. Finally, combining a numerical bound on an OPE coefficient and some
Ward identities involving the current and the displacement operator, we bound
the displacement operator two-point function above. This upper bound also
constrains a boundary contribution to the anomaly in the trace of the stress
tensor for these types of theories.
| arxiv topic:hep-th |
arxiv_dataset-196962312.07792 | Differentially private projection-depth-based medians
math.ST cs.CR cs.LG stat.ME stat.TH
We develop $(\epsilon,\delta)$-differentially private projection-depth-based
medians using the propose-test-release (PTR) and exponential mechanisms. Under
general conditions on the input parameters and the population measure, (e.g. we
do not assume any moment bounds), we quantify the probability the test in PTR
fails, as well as the cost of privacy via finite sample deviation bounds. Next,
we show that when some observations are contaminated, the private
projection-depth-based median does not break down, provided its input location
and scale estimators do not break down. We demonstrate our main results on the
canonical projection-depth-based median, as well as on projection-depth-based
medians derived from trimmed estimators. In the Gaussian setting, we show that
the resulting deviation bound matches the known lower bound for private
Gaussian mean estimation. In the Cauchy setting, we show that the ``outlier
error amplification'' effect resulting from the heavy tails outweighs the cost
of privacy. This result is then verified via numerical simulations.
Additionally, we present results on general PTR mechanisms and a uniform
concentration result on the projected spacings of order statistics, which may
be of general interest.
| arxiv topic:math.ST cs.CR cs.LG stat.ME stat.TH |
arxiv_dataset-196972312.07892 | PT-symmetric quantum sensing: advantages and restrictions
quant-ph
Quantum sensing utilizing unique quantum properties of non-Hermitian systems
to realize ultra-precision measurements has been attracting increasing
attention. However, the debate on whether non-Hermitian systems are superior to
Hermitian counterparts in sensing remains an open question. Here, we
investigate the quantum information in PT-symmetric quantum sensing utilizing
two experimental schemes based on the trapped-ion platform. It turns out that
the existence of advantages of non-Hermitian quantum sensing heavily depends on
additional information resources carried by the extra degrees of freedom
introduced to construct PT-symmetric quantum sensors. Moreover, the practical
application of non-Hermitian quantum sensing with superior performance is
primarily restricted by the additional resource consumption accompanied by the
post-selection. Our study provides theoretical references for the construction
of non-Hermitian quantum sensors with superior performance and has potential
applications in research fields of quantum precision measurement and quantum
information processing.
| arxiv topic:quant-ph |
arxiv_dataset-196982312.07992 | On the privacy of federated Clustering: A Cryptographic View
cs.CR
The privacy concern in federated clustering has attracted considerable
attention in past decades. Many privacy-preserving clustering algorithms
leverage cryptographic techniques like homomorphic encryption or secure
multiparty computation, to guarantee full privacy, i.e., no additional
information is leaked other than the final output. However, given the iterative
nature of clustering algorithms, consistently encrypting intermediate outputs,
such as centroids, hampers efficiency. This paper delves into this intricate
trade-off, questioning the necessity of continuous encryption in iterative
algorithms. Using the federated K-means clustering as an example, we
mathematically formulate the problem of reconstructing input private data from
the intermediate centroids as a classical cryptographic problem called hidden
subset sum problem (HSSP)-extended from an NP-complete problem called subset
sum problem (SSP). Through an in-depth analysis, we show that existing
lattice-based HSSP attacks fail in reconstructing the private data given the
knowledge of intermediate centroids, thus it is secure to reveal them for the
sake of efficiency. To the best of our knowledge, our work is the first to cast
federated clustering's privacy concerns as a cryptographic problem HSSP such
that a concrete and rigorous analysis can be conducted.
| arxiv topic:cs.CR |
arxiv_dataset-196992312.08092 | A hybrid analysis of LBSN data to early detect anomalies in crowd
dynamics
cs.SI cs.AI
Undoubtedly, Location-based Social Networks (LBSNs) provide an interesting
source of geo-located data that we have previously used to obtain patterns of
the dynamics of crowds throughout urban areas. According to our previous
results, activity in LBSNs reflects the real activity in the city. Therefore,
unexpected behaviors in the social media activity are a trustful evidence of
unexpected changes of the activity in the city. In this paper we introduce a
hybrid solution to early detect these changes based on applying a combination
of two approaches, the use of entropy analysis and clustering techniques, on
the data gathered from LBSNs. In particular, we have performed our experiments
over a data set collected from Instagram for seven months in New York City,
obtaining promising results.
| arxiv topic:cs.SI cs.AI |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.