arXiv subjects

Chang Liu

Publications and source records attributed to Chang Liu.

At least 631 records · Page 35Linked to original sources

arXiv2024

A Systematic Review of Deep Learning-based Research on Radiology Report Generation

Radiology report generation (RRG) aims to automatically generate free-text descriptions from clinical radiographs, e.g., chest X-Ray images. RRG plays an essential role in promoting clinical automation and presents significant help to provide practical assistance for inexperienced doctors and alleviate radiologists' workloads. Therefore, consider these meaningful potentials, research on RRG is experiencing explosive growth in the past half-decade, especially with the rapid development of deep learning approaches. Existing studies perform RRG from the perspective of enhancing different modalities, provide insights on optimizing the report generation process with elaborated features from both visual and textual information, and further facilitate RRG with the cross-modal interactions among them. In this paper, we present a comprehensive review of deep learning-based RRG from various perspectives. Specifically, we firstly cover pivotal RRG approaches based on the task-specific features of radiographs, reports, and the cross-modal relations between them, and then illustrate the benchmark datasets conventionally used for this task with evaluation metrics, subsequently analyze the performance of different approaches and finally offer our summary on the challenges and the trends in future directions. Overall, the goal of this paper is to serve as a tool for understanding existing literature and inspiring potential valuable research in the field of RRG.

cs.CV↗

arXiv2024

Fluctuation-Induced Transitions in Anisotropic Two-Dimensional Turbulence

Two-dimensional (2D) turbulence features an inverse energy cascade that produces large-scale flow structures such as large-scale vortices (LSVs) and unidirectional jets. We investigate the dynamics of such structures using extensive direct numerical simulations (DNS) of randomly forced, viscously damped 2D turbulence within a periodic rectangular (Cartesian) domain $[0,L_x]\times[0,L_y]$. LSVs form and dominate the system when the domain aspect ratio $δ= L_x/L_y \approx 1$, while unidirectional jets predominate at $δ\gtrsim 1.1$. At intermediate $δ$, both structures are metastable, with noise-induced transitions between LSVs and jets. We derive and verify predictions for the dependence of kinetic energy E and flow polarity on the nondimensional control parameters. We further collect detailed statistics on the lifetimes of LSVs and jets from DNS runs that are up to 10738 viscous diffusive times long. The distribution of the lifetimes is consistent with that of a memoryless process. Our DNS show an exponential dependence of the mean lifetime on $δ$. Mean lifetimes depend sensitively on the Reynolds number Re: as Re increases, the energy gap between LSV (lower E) and jet states (higher E) arising from anisotropic dissipation increases, leading to an approximately exponential increase in lifetimes with Re for both LSVs and jets. Similarly, as the forcing scale decreases, transitions become less frequent. We study the transitions in detail, revealing that they occur in two stages: an initial, rapid redistribution of kinetic energy by nonlinear triadic interactions deforms LSVs into jets or vice versa. In the second stage, the energy of the newly formed structure slowly adjusts to its associated equilibrium value on a longer, viscous timescale, producing hysteresis. Our findings shed new light on the dynamics of coherent large-scale structures in anisotropic turbulence.

physics.flu-dyn↗

arXiv2024

Change-point analysis for binomial autoregressive model with application to price stability counts

The first-order binomial autoregressive (BAR(1)) model is the most frequently used tool to analyze the bounded count time series. The BAR(1) model is stationary and assumes process parameters to remain constant throughout the time period, which may be incompatible with the non-stationary real data, which indicates piecewise stationary characteristic. To better analyze the non-stationary bounded count time series, this article introduces the BAR(1) process with multiple change-points, which contains the BAR(1) model as a special case. Our primary goals are not only to detect the change-points, but also to give a solution to estimate the number and locations of the change-points. For this, the cumulative sum (CUSUM) test and minimum description length (MDL) principle are employed to deal with the testing and estimation problems. The proposed approaches are also applied to analysis of the Harmonised Index of Consumer Prices of the European Union.

stat.ME↗

arXiv2024

Fabrication-induced even-odd discrepancy of magnetotransport in few-layer MnBi$_2$Te$_4$

The van der Waals antiferromagnetic topological insulator MnBi2Te4 represents a promising platform for exploring the layer-dependent magnetism and topological states of matter. Recently observed discrepancies between magnetic and transport properties have aroused controversies concerning the topological nature of MnBi2Te4 in the ground state. In this article, we demonstrate that fabrication can induce mismatched even-odd layer dependent magnetotransport in few-layer MnBi2Te4. We perform a comprehensive study of the magnetotransport properties in 6- and 7-septuple-layer MnBi2Te4, and reveal that both even- and odd-number-layer device can show zero Hall plateau phenomena in zero magnetic field. Importantly, a statistical survey of the optical contrast in more than 200 MnBi2Te4 flakes reveals that the zero Hall plateau in odd-number-layer devices arises from the reduction of the effective thickness during the fabrication, a factor that was rarely noticed in previous studies of 2D materials. Our finding not only provides an explanation to the controversies regarding the discrepancy of the even-odd layer dependent magnetotransport in MnBi2Te4, but also highlights the critical issues concerning the fabrication and characterization of 2D material devices.

cond-mat.mes-hall↗

arXiv2024

Consisaug: A Consistency-based Augmentation for Polyp Detection in Endoscopy Image Analysis

Colorectal cancer (CRC), which frequently originates from initially benign polyps, remains a significant contributor to global cancer-related mortality. Early and accurate detection of these polyps via colonoscopy is crucial for CRC prevention. However, traditional colonoscopy methods depend heavily on the operator's experience, leading to suboptimal polyp detection rates. Besides, the public database are limited in polyp size and shape diversity. To enhance the available data for polyp detection, we introduce Consisaug, an innovative and effective methodology to augment data that leverages deep learning. We utilize the constraint that when the image is flipped the class label should be equal and the bonding boxes should be consistent. We implement our Consisaug on five public polyp datasets and at three backbones, and the results show the effectiveness of our method.

cs.CV↗

arXiv2024

Constraints on Einstein-dilaton Gauss-Bonnet gravity with Taiji

In the 2030s, space-based gravitational-wave (GW) detectors will exhibit unprecedented sensitivity in the millihertz frequency band, greatly expanding the potential for testing theories of gravity compared to ground-based GW detectors. Inspired by effective string theory, Einstein-dilaton Gauss-Bonnet (EdGB) gravity introduces an extra dilaton scalar field that is directly coupled to higher curvature terms. Here, we investigate the capability of Taiji to constrain the parameters of EdGB gravity by analyzing GWs from massive black hole binaries (MBHBs). We utilize the parameterized post-Einsteinian (ppE) waveform with the leading order EdGB corrections for the inspiral phase of MBHBs. The constraints on the coupling constants are obtained by performing Fisher matrix analysis. With different mass ratios and spins $χ_i$ at redshifts $z=2,3,4,5$, the $1σ$ bounds on the parameter $α$ have the same order of magnitude: $\sqrtα\sim 10^7$m.

gr-qc↗

arXiv2024

Bifurcation delay and front propagation in the real Ginzburg-Landau equation on a time-dependent domain

This work analyzes bifurcation delay and front propagation in the one-dimensional real Ginzburg-Landau equation (RGLE) with periodic boundary conditions on monotonically growing or shrinking domains. First, we obtain closed-form expressions for the delay of primary bifurcations on a growing domain and show that the additional domain growth before the appearance of a pattern is independent of the growth time scale. We also quantify primary bifurcation delay on a shrinking domain; the time scale of domain compression is reflected in the additional compression before the pattern decays. For secondary bifurcations such as the Eckhaus instability, we obtain a lower bound on the delay of phase slips due to a time-dependent domain. We also construct a heuristic model to classify regimes with arrested phase slips, i.e. phase slips that fail to develop. Then, we study how fronts are influenced by a time-dependent domain. We derive expressions for the velocity and profile of homogeneous fronts on a time-dependent domain. We also derive the natural ``asymptotic'' velocity and front profile and show that these deviate from predictions based on the marginal stability criterion familiar from fixed domain theory. This difference arises because the time-dependence of the domain lifts the degeneracy of the spatial eigenvalues associated with speed selection and represents a fundamental distinction from the fixed domain theory that we verify using direct numerical simulations. The effect of a growing domain on pattern-spreading and Eckhaus front velocities is inspected qualitatively and found to be similar to that of homogeneous fronts. These more complex fronts can also experience delayed onset. Lastly, we show that dilution -- an effect present when the order parameter is conserved -- increases bifurcation delay and amplifies changes in the homogeneous front velocity on time-dependent domains.

nlin.PS↗

arXiv2024

Convolutional Bayesian Filtering

Bayesian filtering serves as the mainstream framework of state estimation in dynamic systems. Its standard version utilizes total probability rule and Bayes' law alternatively, where how to define and compute conditional probability is critical to state distribution inference. Previously, the conditional probability is assumed to be exactly known, which represents a measure of the occurrence probability of one event, given the second event. In this paper, we find that by adding an additional event that stipulates an inequality condition, we can transform the conditional probability into a special integration that is analogous to convolution. Based on this transformation, we show that both transition probability and output probability can be generalized to convolutional forms, resulting in a more general filtering framework that we call convolutional Bayesian filtering. This new framework encompasses standard Bayesian filtering as a special case when the distance metric of the inequality condition is selected as Dirac delta function. It also allows for a more nuanced consideration of model mismatch by choosing different types of inequality conditions. For instance, when the distance metric is defined in a distributional sense, the transition probability and output probability can be approximated by simply rescaling them into fractional powers. Under this framework, a robust version of Kalman filter can be constructed by only altering the noise covariance matrix, while maintaining the conjugate nature of Gaussian distributions. Finally, we exemplify the effectiveness of our approach by reshaping classic filtering algorithms into convolutional versions, including Kalman filter, extended Kalman filter, unscented Kalman filter and particle filter.

stat.ML↗

arXiv2024

Resource Allocation in Large Language Model Integrated 6G Vehicular Networks

In the upcoming 6G era, vehicular networks are shifting from simple Vehicle-to-Vehicle (V2V) communication to the more complex Vehicle-to-Everything (V2X) connectivity. At the forefront of this shift is the incorporation of Large Language Models (LLMs) into vehicles. Known for their sophisticated natural language processing abilities, LLMs change how users interact with their vehicles. This integration facilitates voice-driven commands and interactions, departing from the conventional manual control systems. However, integrating LLMs into vehicular systems presents notable challenges. The substantial computational demands and energy requirements of LLMs pose significant challenges, especially in the constrained environment of a vehicle. Additionally, the time-sensitive nature of tasks in vehicular networks adds another layer of complexity. In this paper, we consider an edge computing system where vehicles process the initial layers of LLM computations locally, and offload the remaining LLM computation tasks to the Roadside Units (RSUs), envisioning a vehicular ecosystem where LLM computations seamlessly interact with the ultra-low latency and high-bandwidth capabilities of 6G networks. To balance the trade-off between completion time and energy consumption, we formulate a multi-objective optimization problem to minimize the total cost of the vehicles and RSUs. The problem is then decomposed into two sub-problems, which are solved by sequential quadratic programming (SQP) method and fractional programming technique. The simulation results clearly indicate that the algorithm we have proposed is highly effective in reducing both the completion time and energy consumption of the system.

cs.DC↗

arXiv2024

Offloading and Quality Control for AI Generated Content Services in 6G Mobile Edge Computing Networks

AI-Generated Content (AIGC), as a novel manner of providing Metaverse services in the forthcoming Internet paradigm, can resolve the obstacles of immersion requirements. Concurrently, edge computing, as an evolutionary paradigm of computing in communication systems, effectively augments real-time interactive services. In pursuit of enhancing the accessibility of AIGC services, the deployment of AIGC models (e.g., diffusion models) to edge servers and local devices has become a prevailing trend. Nevertheless, this approach faces constraints imposed by battery life and computational resources when tasks are offloaded to local devices, limiting the capacity to deliver high-quality content to users while adhering to stringent latency requirements. So there will be a tradeoff between the utility of AIGC models and offloading decisions in the edge computing paradigm. This paper proposes a joint optimization algorithm for offloading decisions, computation time, and diffusion steps of the diffusion models in the reverse diffusion stage. Moreover, we take the average error into consideration as the metric for evaluating the quality of the generated results. Experimental results conclusively demonstrate that the proposed algorithm achieves superior joint optimization performance compared to the baselines.

cs.AI↗

arXiv2024

Nanoscale Imaging of Phonons and Reconfiguration in Topologically-Engineered, Self-Assembled Nanoparticle Lattice

Topologically-engineered mechanical frames are important model constructs for architecture, machine mechanisms, and metamaterials. Despite significant advances in macroscopically fashioned frames, realization and phonon imaging of nanoframes have remained challenging. Here we extend for the first time the principles of topologically-engineered mechanical frames to lattices self-assembled from nanoparticles. Liquid-phase transmission electron microscopy images the vibrations of nanoparticles in self-assembled Maxwell and hexagonal lattices at the nanometer resolution, measuring a series of otherwise inaccessible properties such as phonon spectra and nonlinear lattice deformation paths. These properties are experimentally modulated by ionic strength, captured by our discrete mechanical model considering the complexity of nanoscale interactions and thermal fluctuations. The experiment-theory integration bridges mechanical metamaterials and colloidal self-assembly, opening new opportunities to manufacture phononic devices with solution processibility, transformability, light weight, and emergent functions, at underexplored length, frequency, and energy scales.

cond-mat.mtrl-sci↗

arXiv2024

Weakly interacting Bose gas with two-body losses

We study the many-body dynamics of weakly interacting Bose gases with two-particle losses. We show that both the two-body interactions and losses in atomic gases may be tuned by controlling the inelastic scattering process between atoms by an optical Feshbach resonance. Interestingly, the low-energy behavior of the scattering amplitude is governed by a single parameter, i.e. the complex $s$-wave scattering length $a_c$. The many-body dynamics are thus described by a Lindblad master equation with complex scattering length. We solve this equation by applying the Bogoliubov approximation in analogy to the closed systems. Various peculiar dynamical properties are discovered, some of them may be regarded as the dissipative counterparts of the celebrated results in closed Bose gases. For example, we show that the next-order correction to the mean-field particle decay rate is to the order of $|n a_c^{3}|^{1/2}$, which is an analogy of the Lee-Huang-Yang correction of Bose gases. It is also found that there exists a dynamical symmetry of symplectic group Sp$(4,\mathbb{C})$ in the quadratic Bogoliubov master equation, which is an analogy of the SU(1,1) dynamical symmetry of the corresponding closed system. We further confirmed the validity of the Bogoliubov approximation by comparing its results with a full numerical calculation in a double-well toy model. Generalizations of other alternative approaches such as the dissipative version of the Gross-Pitaevskii equation and hydrodynamic theory are also discussed in the last.

cond-mat.quant-gas↗

arXiv2024

Space-Efficient Indexes for Uncertain Strings

Strings in the real world are often encoded with some level of uncertainty. In the character-level uncertainty model, an uncertain string $X$ of length $n$ on an alphabet $Σ$ is a sequence of $n$ probability distributions over $Σ$. Given an uncertain string $X$ and a weight threshold $\frac{1}{z}\in(0,1]$, we say that pattern $P$ occurs in $X$ at position $i$, if the product of probabilities of the letters of $P$ at positions $i,\ldots,i+|P|-1$ is at least $\frac{1}{z}$. While indexing standard strings for online pattern searches can be performed in linear time and space, indexing uncertain strings is much more challenging. Specifically, the state-of-the-art index for uncertain strings has $\mathcal{O}(nz)$ size, requires $\mathcal{O}(nz)$ time and $\mathcal{O}(nz)$ space to be constructed, and answers pattern matching queries in the optimal $\mathcal{O}(m+|\text{Occ}|)$ time, where $m$ is the length of $P$ and $|\text{Occ}|$ is the total number of occurrences of $P$ in $X$. For large $n$ and (moderate) $z$ values, this index is completely impractical to construct, which outweighs the benefit of the supported optimal pattern matching queries. We were thus motivated to design a space-efficient index at the expense of slower yet competitive pattern matching queries. We propose an index of $\mathcal{O}(\frac{nz}{\ell}\log z)$ expected size, which can be constructed using $\mathcal{O}(\frac{nz}{\ell}\log z)$ expected space, and supports very fast pattern matching queries in expectation, for patterns of length $m\geq \ell$. We have implemented and evaluated several versions of our index. The best-performing version of our index is up to two orders of magnitude smaller than the state of the art in terms of both index size and construction space, while offering faster or very competitive query and construction times.

cs.DS↗

arXiv2024

RGBGrasp: Image-based Object Grasping by Capturing Multiple Views during Robot Arm Movement with Neural Radiance Fields

Robotic research encounters a significant hurdle when it comes to the intricate task of grasping objects that come in various shapes, materials, and textures. Unlike many prior investigations that heavily leaned on specialized point-cloud cameras or abundant RGB visual data to gather 3D insights for object-grasping missions, this paper introduces a pioneering approach called RGBGrasp. This method depends on a limited set of RGB views to perceive the 3D surroundings containing transparent and specular objects and achieve accurate grasping. Our method utilizes pre-trained depth prediction models to establish geometry constraints, enabling precise 3D structure estimation, even under limited view conditions. Finally, we integrate hash encoding and a proposal sampler strategy to significantly accelerate the 3D reconstruction process. These innovations significantly enhance the adaptability and effectiveness of our algorithm in real-world scenarios. Through comprehensive experimental validations, we demonstrate that RGBGrasp achieves remarkable success across a wide spectrum of object-grasping scenarios, establishing it as a promising solution for real-world robotic manipulation tasks. The demonstrations of our method can be found on: https://sites.google.com/view/rgbgrasp

cs.RO↗

arXiv2024

Prospects for Taiji to detect a gravitational-wave background from cosmic strings

Recently, multiple pulsar timing array collaborations have presented compelling evidence for a stochastic signal at nanohertz frequencies, potentially originating from cosmic strings. Cosmic strings are linear topological defects that can arise during phase transitions in the early Universe or as fundamental strings in superstring theory. This paper focuses on investigating the detection capabilities of Taiji, a planned space-based gravitational wave detector, for the gravitational wave background generated by cosmic strings. By analyzing simulated Taiji data and utilizing comprehensive Bayesian parameter estimation techniques, we demonstrate a significant improvement in precision compared to the NANOGrav 15-year data, surpassing it by an order of magnitude. This highlights the enhanced measurement capabilities of Taiji. Consequently, Taiji can serve as a valuable complementary tool to pulsar timing arrays in validating and exploring the physics of cosmic strings in the early Universe.

astro-ph.IM↗

arXiv2024

FaceChain-SuDe: Building Derived Class to Inherit Category Attributes for One-shot Subject-Driven Generation

Subject-driven generation has garnered significant interest recently due to its ability to personalize text-to-image generation. Typical works focus on learning the new subject's private attributes. However, an important fact has not been taken seriously that a subject is not an isolated new concept but should be a specialization of a certain category in the pre-trained model. This results in the subject failing to comprehensively inherit the attributes in its category, causing poor attribute-related generations. In this paper, motivated by object-oriented programming, we model the subject as a derived class whose base class is its semantic category. This modeling enables the subject to inherit public attributes from its category while learning its private attributes from the user-provided example. Specifically, we propose a plug-and-play method, Subject-Derived regularization (SuDe). It constructs the base-derived class modeling by constraining the subject-driven generated images to semantically belong to the subject's category. Extensive experiments under three baselines and two backbones on various subjects show that our SuDe enables imaginative attribute-related generations while maintaining subject fidelity. Codes will be open sourced soon at FaceChain (https://github.com/modelscope/facechain).

cs.CV↗

arXiv2024

Overcoming the Barrier of Orbital-Free Density Functional Theory for Molecular Systems Using Deep Learning

Orbital-free density functional theory (OFDFT) is a quantum chemistry formulation that has a lower cost scaling than the prevailing Kohn-Sham DFT, which is increasingly desired for contemporary molecular research. However, its accuracy is limited by the kinetic energy density functional, which is notoriously hard to approximate for non-periodic molecular systems. Here we propose M-OFDFT, an OFDFT approach capable of solving molecular systems using a deep learning functional model. We build the essential non-locality into the model, which is made affordable by the concise density representation as expansion coefficients under an atomic basis. With techniques to address unconventional learning challenges therein, M-OFDFT achieves a comparable accuracy with Kohn-Sham DFT on a wide range of molecules untouched by OFDFT before. More attractively, M-OFDFT extrapolates well to molecules much larger than those seen in training, which unleashes the appealing scaling of OFDFT for studying large molecules including proteins, representing an advancement of the accuracy-efficiency trade-off frontier in quantum chemistry.

stat.ML↗

arXiv2024

Articulated Object Manipulation with Coarse-to-fine Affordance for Mitigating the Effect of Point Cloud Noise

3D articulated objects are inherently challenging for manipulation due to the varied geometries and intricate functionalities associated with articulated objects.Point-level affordance, which predicts the per-point actionable score and thus proposes the best point to interact with, has demonstrated excellent performance and generalization capabilities in articulated object manipulation. However, a significant challenge remains: while previous works use perfect point cloud generated in simulation, the models cannot directly apply to the noisy point cloud in the real-world. To tackle this challenge, we leverage the property of real-world scanned point cloud that, the point cloud becomes less noisy when the camera is closer to the object. Therefore, we propose a novel coarse-to-fine affordance learning pipeline to mitigate the effect of point cloud noise in two stages. In the first stage, we learn the affordance on the noisy far point cloud which includes the whole object to propose the approximated place to manipulate. Then, we move the camera in front of the approximated place, scan a less noisy point cloud containing precise local geometries for manipulation, and learn affordance on such point cloud to propose fine-grained final actions. The proposed method is thoroughly evaluated both using large-scale simulated noisy point clouds mimicking real-world scans, and in the real world scenarios, with superiority over existing methods, demonstrating the effectiveness in tackling the noisy real-world point cloud problem.

cs.RO↗