arXiv subjects

Cheng Chen

Publications and source records attributed to Cheng Chen.

At least 235 records · Page 13Linked to original sources

arXiv2023

Class Gradient Projection For Continual Learning

Catastrophic forgetting is one of the most critical challenges in Continual Learning (CL). Recent approaches tackle this problem by projecting the gradient update orthogonal to the gradient subspace of existing tasks. While the results are remarkable, those approaches ignore the fact that these calculated gradients are not guaranteed to be orthogonal to the gradient subspace of each class due to the class deviation in tasks, e.g., distinguishing "Man" from "Sea" v.s. differentiating "Boy" from "Girl". Therefore, this strategy may still cause catastrophic forgetting for some classes. In this paper, we propose Class Gradient Projection (CGP), which calculates the gradient subspace from individual classes rather than tasks. Gradient update orthogonal to the gradient subspace of existing classes can be effectively utilized to minimize interference from other classes. To improve the generalization and efficiency, we further design a Base Refining (BR) algorithm to combine similar classes and refine class bases dynamically. Moreover, we leverage a contrastive learning method to improve the model's ability to handle unseen tasks. Extensive experiments on benchmark datasets demonstrate the effectiveness of our proposed approach. It improves the previous methods by 2.0% on the CIFAR-100 dataset.

cs.CV↗

arXiv2023

Program Dependence Net and On-demand Slicing for Property Verification of Concurrent System and Software

When checking concurrent software using a finite-state model, we face a formidable state explosion problem. One solution to this problem is dependence-based program slicing, whose use can effectively reduce verification time. It is orthogonal to other model-checking reduction techniques. However, when slicing concurrent programs for model checking, there are conversions between multiple irreplaceable models, and dependencies need to be found for variables irrelevant to the verified property, which results in redundant computation. To resolve this issue, we propose a Program Dependence Net (PDNet) based on Petri net theory. It is a unified model that combines a control-flow structure with dependencies to avoid conversions. For reduction, we present a PDNet slicing method to capture the relevant variables' dependencies when needed. PDNet in verifying linear temporal logic and its on-demand slicing can be used to significantly reduce computation cost. We implement a model-checking tool based on PDNet and its on-demand slicing, and validate the advantages of our proposed methods.

cs.SE↗

arXiv2023

Nash's Existence Theorem for Non-compact Strategy Sets

This paper generalizes the Fan-Knaster-Kuratowski-Mazurkiewicz (FKKM) lemma to the case of weak topology, and obtains the Ky Fan minimax inequality defined on non-empty non-compact convex subsets in reflexive Banach spaces, then we apply it to game theory and obtain Nash's existence theorem for non-compact strategy sets, together with John von Neumann's existence theorem in two-player zero-sum games.

math.FA↗

arXiv2023

Building Real-World Meeting Summarization Systems using Large Language Models: A Practical Perspective

This paper studies how to effectively build meeting summarization systems for real-world usage using large language models (LLMs). For this purpose, we conduct an extensive evaluation and comparison of various closed-source and open-source LLMs, namely, GPT-4, GPT- 3.5, PaLM-2, and LLaMA-2. Our findings reveal that most closed-source LLMs are generally better in terms of performance. However, much smaller open-source models like LLaMA- 2 (7B and 13B) could still achieve performance comparable to the large closed-source models even in zero-shot scenarios. Considering the privacy concerns of closed-source models for only being accessible via API, alongside the high cost associated with using fine-tuned versions of the closed-source models, the opensource models that can achieve competitive performance are more advantageous for industrial use. Balancing performance with associated costs and privacy concerns, the LLaMA-2-7B model looks more promising for industrial usage. In sum, this paper offers practical insights on using LLMs for real-world business meeting summarization, shedding light on the trade-offs between performance and cost.

cs.CL↗

arXiv2023

Uncertainty Estimation for Safety-critical Scene Segmentation via Fine-grained Reward Maximization

Uncertainty estimation plays an important role for future reliable deployment of deep segmentation models in safety-critical scenarios such as medical applications. However, existing methods for uncertainty estimation have been limited by the lack of explicit guidance for calibrating the prediction risk and model confidence. In this work, we propose a novel fine-grained reward maximization (FGRM) framework, to address uncertainty estimation by directly utilizing an uncertainty metric related reward function with a reinforcement learning based model tuning algorithm. This would benefit the model uncertainty estimation through direct optimization guidance for model calibration. Specifically, our method designs a new uncertainty estimation reward function using the calibration metric, which is maximized to fine-tune an evidential learning pre-trained segmentation model for calibrating prediction risk. Importantly, we innovate an effective fine-grained parameter update scheme, which imposes fine-grained reward-weighting of each network parameter according to the parameter importance quantified by the fisher information matrix. To the best of our knowledge, this is the first work exploring reward optimization for model uncertainty estimation in safety-critical vision tasks. The effectiveness of our method is demonstrated on two large safety-critical surgical scene segmentation datasets under two different uncertainty estimation settings. With real-time one forward pass at inference, our method outperforms state-of-the-art methods by a clear margin on all the calibration metrics of uncertainty estimation, while maintaining a high task accuracy for the segmentation results. Code is available at \url{https://github.com/med-air/FGRM}.

cs.CV↗

arXiv2023

Adversarial Attacks on Cooperative Multi-agent Bandits

Cooperative multi-agent multi-armed bandits (CMA2B) consider the collaborative efforts of multiple agents in a shared multi-armed bandit game. We study latent vulnerabilities exposed by this collaboration and consider adversarial attacks on a few agents with the goal of influencing the decisions of the rest. More specifically, we study adversarial attacks on CMA2B in both homogeneous settings, where agents operate with the same arm set, and heterogeneous settings, where agents have distinct arm sets. In the homogeneous setting, we propose attack strategies that, by targeting just one agent, convince all agents to select a particular target arm $T-o(T)$ times while incurring $o(T)$ attack costs in $T$ rounds. In the heterogeneous setting, we prove that a target arm attack requires linear attack costs and propose attack strategies that can force a maximum number of agents to suffer linear regrets while incurring sublinear costs and only manipulating the observations of a few target agents. Numerical experiments validate the effectiveness of our proposed attack strategies.

cs.LG↗

arXiv2023

Are Large Language Models Reliable Judges? A Study on the Factuality Evaluation Capabilities of LLMs

In recent years, Large Language Models (LLMs) have gained immense attention due to their notable emergent capabilities, surpassing those seen in earlier language models. A particularly intriguing application of LLMs is their role as evaluators for texts produced by various generative models. In this study, we delve into the potential of LLMs as reliable assessors of factual consistency in summaries generated by text-generation models. Initially, we introduce an innovative approach for factuality assessment using LLMs. This entails employing a singular LLM for the entirety of the question-answering-based factuality scoring process. Following this, we examine the efficacy of various LLMs in direct factuality scoring, benchmarking them against traditional measures and human annotations. Contrary to initial expectations, our results indicate a lack of significant correlations between factuality metrics and human evaluations, specifically for GPT-4 and PaLM-2. Notable correlations were only observed with GPT-3.5 across two factuality subcategories. These consistent findings across various factual error categories suggest a fundamental limitation in the current LLMs' capability to accurately gauge factuality. This version presents the information more concisely while maintaining the main points and findings of the original text.

cs.CL↗

arXiv2023

Single-Photon Counting Receivers for Optical Wireless Communications in Future 6G Networks

Optical wireless communication (OWC) offers several complementary advantages to radio-frequency wireless networks such as its massive available spectrum; hence, it is widely anticipated that OWC will assume a pivotal role in the forthcoming sixth generation wireless communication networks. Although significant progress has been achieved in OWC over the past decades, the outage induced by occasionally low received optical power continues to pose a key limiting factor for its deployment. In this work, we discuss the potential role of single-photon counting (SPC) receivers as a promising solution to overcome this limitation. We present an overview of the applications of SPC-based OWC systems in 6G networks, introduce their major performance-limiting factors, propose a performance enhancement framework to tackle these issues, and identify critical areas of open problems for future research.

cs.IT↗

arXiv2023

Tilted circumbinary planetary systems as efficient progenitors of free-floating planets

The dominant mechanism for generating free-floating planets has so far remained elusive. One suggested mechanism is that planets are ejected from planetary systems due to planet-planet interactions. However, instability around a single star requires a very compactly spaced planetary system. We find that around binary star systems instability can occur even with widely separated planets that are on tilted orbits relative to the binary orbit due to combined effects of planet-binary and planet-planet interactions, especially if the binary is on an eccentric orbit. We investigate the orbital stability of planetary systems with various planet masses and architectures. We find that the stability of the system depends upon the mass of the highest mass planet. The order of the planets in the system does not significantly affect stability but, generally, the most massive planet remains stable and the lower mass planets are ejected. The minimum planet mass required to trigger the instability is about that of Neptune for a circular orbit binary and a super-Earth of about $10$ Earth masses for highly eccentric binaries. Hence, we suggest that planet formation around misaligned binaries can be an efficient formation mechanism for free-floating planets. While most observed free-floating planets are giant planets, we predict that there should be more low-mass free floating planets that are as yet unobserved than higher mass planets.

astro-ph.EP↗

arXiv2023

Arithmetic Branching Law and generic $L$-packets

Let $G$ be a classical group defined over a local field $F$ of characteristic zero. For any irreducible admissible representation $π$ of $G(F)$, which is of Casselman-Wallach type if $F$ is archimedean, we extend the study of spectral decomposition of local descents in [JZ18] for special orthogonal groups over non-archimedean local fields to more general classical groups over any local field $F$. In particular, if $π$ has a generic local $L$-parameter, we introduce the spectral first occurrence index $\mathfrak{f}_{\mathfrak{s}}(π)$ and the arithmetic first occurrence index $\mathfrak{f}_{\mathfrak{a}}(π)$ of $π$ and prove in Theorem 1.4 that $\mathfrak{f}_{\mathfrak{s}}(π) = \mathfrak{f}_{\mathfrak{a}}(π)$. Based on the theory of consecutive descents of enhanced $L$-parameters developed in [JLZ22], we are able to show in Theorem 1.5 that the first descent spectrum consists of all discrete series representations, which determines explicitly the branching decomposition problem by means of the relevant arithmetic data and extends the main result ([JZ18, Theorem 1.7]) to the great generality.

math.RT↗

arXiv2023

MA-SAM: Modality-agnostic SAM Adaptation for 3D Medical Image Segmentation

The Segment Anything Model (SAM), a foundation model for general image segmentation, has demonstrated impressive zero-shot performance across numerous natural image segmentation tasks. However, SAM's performance significantly declines when applied to medical images, primarily due to the substantial disparity between natural and medical image domains. To effectively adapt SAM to medical images, it is important to incorporate critical third-dimensional information, i.e., volumetric or temporal knowledge, during fine-tuning. Simultaneously, we aim to harness SAM's pre-trained weights within its original 2D backbone to the fullest extent. In this paper, we introduce a modality-agnostic SAM adaptation framework, named as MA-SAM, that is applicable to various volumetric and video medical data. Our method roots in the parameter-efficient fine-tuning strategy to update only a small portion of weight increments while preserving the majority of SAM's pre-trained weights. By injecting a series of 3D adapters into the transformer blocks of the image encoder, our method enables the pre-trained 2D backbone to extract third-dimensional information from input data. The effectiveness of our method has been comprehensively evaluated on four medical image segmentation tasks, by using 10 public datasets across CT, MRI, and surgical video data. Remarkably, without using any prompt, our method consistently outperforms various state-of-the-art 3D approaches, surpassing nnU-Net by 0.9%, 2.6%, and 9.9% in Dice for CT multi-organ segmentation, MRI prostate segmentation, and surgical scene segmentation respectively. Our model also demonstrates strong generalization, and excels in challenging tumor segmentation when prompts are used. Our code is available at: https://github.com/cchen-cc/MA-SAM.

cs.CV↗

arXiv2023

Selection of contributing factors for predicting landslide susceptibility using machine learning and deep learning models

Landslides are a common natural disaster that can cause casualties, property safety threats and economic losses. Therefore, it is important to understand or predict the probability of landslide occurrence at potentially risky sites. A commonly used means is to carry out a landslide susceptibility assessment based on a landslide inventory and a set of landslide contributing factors. This can be readily achieved using machine learning (ML) models such as logistic regression (LR), support vector machine (SVM), random forest (RF), extreme gradient boosting (Xgboost), or deep learning (DL) models such as convolutional neural network (CNN) and long short time memory (LSTM). As the input data for these models, landslide contributing factors have varying influences on landslide occurrence. Therefore, it is logically feasible to select more important contributing factors and eliminate less relevant ones, with the aim of increasing the prediction accuracy of these models. However, selecting more important factors is still a challenging task and there is no generally accepted method. Furthermore, the effects of factor selection using various methods on the prediction accuracy of ML and DL models are unclear. In this study, the impact of the selection of contributing factors on the accuracy of landslide susceptibility predictions using ML and DL models was investigated. Four methods for selecting contributing factors were considered for all the aforementioned ML and DL models, which included Information Gain Ratio (IGR), Recursive Feature Elimination (RFE), Particle Swarm Optimization (PSO), Least Absolute Shrinkage and Selection Operators (LASSO) and Harris Hawk Optimization (HHO). In addition, autoencoder-based factor selection methods for DL models were also investigated. To assess their performances, an exhaustive approach was adopted,...

cs.LG↗

arXiv2023

Anomalous excitonic phase diagram in band-gap-tuned Ta2Ni(Se,S)5

During a band-gap-tuned semimetal-to-semiconductor transition, Coulomb attraction between electrons and holes can cause spontaneously formed excitons near the zero-band-gap point, or the Lifshitz transition point. This has become an important route to realize bulk excitonic insulators -- an insulating ground state distinct from single-particle band insulators. How this route manifests from weak to strong coupling is not clear. In this work, using angle-resolved photoemission spectroscopy (ARPES) and high-resolution synchrotron x-ray diffraction (XRD), we investigate the broken symmetry state across the semimetal-to-semiconductor transition in a leading bulk excitonic insulator candidate system Ta2Ni(Se,S)5. A broken symmetry phase is found to be continuously suppressed from the semimetal side to the semiconductor side, contradicting the anticipated maximal excitonic instability around the Lifshitz transition. Bolstered by first-principles and model calculations, we find strong interband electron-phonon coupling to play a crucial role in the enhanced symmetry breaking on the semimetal side of the phase diagram. Our results not only provide insight into the longstanding debate of the nature of intertwined orders in Ta2NiSe5, but also establish a basis for exploring band-gap-tuned structural and electronic instabilities in strongly coupled systems.

cond-mat.str-el↗

arXiv2023

Scalable spin squeezing in a dipolar Rydberg atom array

The standard quantum limit bounds the precision of measurements that can be achieved by ensembles of uncorrelated particles. Fundamentally, this limit arises from the non-commuting nature of quantum mechanics, leading to the presence of fluctuations often referred to as quantum projection noise. Quantum metrology relies on the use of non-classical states of many-body systems in order to enhance the precision of measurements beyond the standard quantum limit. To do so, one can reshape the quantum projection noise -- a strategy known as squeezing. In the context of many-body spin systems, one typically utilizes all-to-all interactions (e.g. the one-axis twisting model) between the constituents to generate the structured entanglement characteristic of spin squeezing. Motivated by recent theoretical work, here we explore the prediction that short-range interactions -- and in particular, the two-dimensional dipolar XY model -- can also enable the realization of scalable spin squeezing. Working with a dipolar Rydberg quantum simulator of up to 100 atoms, we demonstrate that quench dynamics from a polarized initial state lead to spin squeezing that improves with increasing system size up to a maximum of -3.5 dB (prior to correcting for detection errors, or approximately -5 dB after correction). Finally, we present two independent refinements: first, using a multistep spin-squeezing protocol allows us to further enhance the squeezing by approximately 1 dB, and second, leveraging Floquet engineering to realize Heisenberg interactions, we demonstrate the ability to extend the lifetime of the squeezed state by freezing its dynamics.

quant-ph↗

arXiv2023

Radiology-Llama2: Best-in-Class Large Language Model for Radiology

This paper introduces Radiology-Llama2, a large language model specialized for radiology through a process known as instruction tuning. Radiology-Llama2 is based on the Llama2 architecture and further trained on a large dataset of radiology reports to generate coherent and clinically useful impressions from radiological findings. Quantitative evaluations using ROUGE metrics on the MIMIC-CXR and OpenI datasets demonstrate that Radiology-Llama2 achieves state-of-the-art performance compared to other generative language models, with a Rouge-1 score of 0.4834 on MIMIC-CXR and 0.4185 on OpenI. Additional assessments by radiology experts highlight the model's strengths in understandability, coherence, relevance, conciseness, and clinical utility. The work illustrates the potential of localized language models designed and tuned for specialized domains like radiology. When properly evaluated and deployed, such models can transform fields like radiology by automating rote tasks and enhancing human expertise.

cs.CL↗

arXiv2023

Can a binary star host three giant circumbinary planets?

We investigate the orbital stability of a tilted circumbinary planetary system with three giant planets. The planets are spaced by a constant number ($Δ$) of mutual Hill radii in the range $Δ=3.4-12.0$ such that the period ratio of the inner pair is the same as the outer pair. A tilted circumbinary planetary system can be unstable even if the same system around a coplanar binary is stable. For an equal mass binary, we find that the stability of a three-planet system is qualitatively similar to that of a two-planet system, but the three-planet system is more unstable in mean motion resonance regions. For an unequal mass binary, there is significantly more instability in the three-planet system as the inner planets can undergo von-Zeipel-Kozai-Lidov oscillations. Generally in unstable systems, the inner planets are more likely to be ejected than the outer planets. The most likely unstable outcome for closely spaced systems, with $Δ\lesssim 8$, is a single remaining stable planet. For more widely separated systems, $Δ\gtrsim 8$, the most likely unstable outcome is two stable planets, only one being ejected. An observed circumbinary planet with significant eccentricity may suggest that it was formed from an unstable system. Consequently, a binary can host three tilted giant planets if the binary stars are close to equal mass and provided that the planets are well spaced and not close to a mean motion resonance.

astro-ph.EP↗

arXiv2023

Multiplicity formula for induced representations: Bessel and Fourier-Jacobi models over Archimedean local fields

This article proves a formula relating the multiplicity of an induced representation and that of the inducing datum for the Bessel and the Fourier-Jacobi models over Archimedean local fields by generalizing the approach of C. Moeglin and J.-L. Waldspurger in [MW12], which was successful for Bessel models of special orthogonal groups over non-Archimedean local fields. As an application, we give a uniform proof of the local Gan-Gross-Prasad conjecture for all classical groups over Archimedean local fields for generic local $L$-parameters based on the tempered basic cases.

math.NT↗

arXiv2023

Treatment Outcome Prediction for Intracerebral Hemorrhage via Generative Prognostic Model with Imaging and Tabular Data

Intracerebral hemorrhage (ICH) is the second most common and deadliest form of stroke. Despite medical advances, predicting treat ment outcomes for ICH remains a challenge. This paper proposes a novel prognostic model that utilizes both imaging and tabular data to predict treatment outcome for ICH. Our model is trained on observational data collected from non-randomized controlled trials, providing reliable predictions of treatment success. Specifically, we propose to employ a variational autoencoder model to generate a low-dimensional prognostic score, which can effectively address the selection bias resulting from the non-randomized controlled trials. Importantly, we develop a variational distributions combination module that combines the information from imaging data, non-imaging clinical data, and treatment assignment to accurately generate the prognostic score. We conducted extensive experiments on a real-world clinical dataset of intracerebral hemorrhage. Our proposed method demonstrates a substantial improvement in treatment outcome prediction compared to existing state-of-the-art approaches. Code is available at https://github.com/med-air/TOP-GPM

cs.CV↗