arXiv subjects

Cheng Li

Publications and source records attributed to Cheng Li.

At least 217 records · Page 12Linked to original sources

arXiv2024

Jupiter's Polar Vortex Crystals Explored using the Shallow Water Equations

At the poles of Jupiter, cyclonic vortices are clustered together in patterns made up of equilateral triangles called vortex crystals. Such patterns are seen in laboratory flows but never before in a planetary atmosphere, where the planet's rotation and gravity add new physics. Here we use the shallow water (SW) equations at the pole of a rotating planet to study the emergence and evolution of vortices starting from an initial random pattern of small-scale turbulence. The flow is in a single layer with a free surface whose slope produces the horizontal pressure gradient force. We explored three parameters in the problem: the mean kinetic energy of the initial turbulence, the horizontal scale of the initial turbulence, and the radius of deformation of the undisturbed fluid layer. We find that some regions of this parameter space lead to vortex crystals and others lead to chaotic behavior and mergers. Our results identified that the relative change of the layer thickness is the key quantity that determines whether the vortex crystal or chaotic patterns emerge.

astro-ph.EP↗

arXiv2024

Enhancing Representation in Medical Vision-Language Foundation Models via Multi-Scale Information Extraction Techniques

The development of medical vision-language foundation models has attracted significant attention in the field of medicine and healthcare due to their promising prospect in various clinical applications. While previous studies have commonly focused on feature learning at a single learning scale, investigation on integrating multi-scale information is lacking, which may hinder the potential for mutual reinforcement among these features. This paper aims to bridge this gap by proposing a method that effectively exploits multi-scale information to enhance the performance of medical foundation models. The proposed method simultaneously exploits features at the local, instance, modality and global aspects, facilitating comprehensive representation learning within the models. We evaluate the effectiveness of the proposed method on six open-source datasets across different clinical tasks, demonstrating its ability to enhance the performance of medical foundation models.

cs.CV↗

arXiv2024

Bridging the Preference Gap between Retrievers and LLMs

Large Language Models (LLMs) have demonstrated superior results across a wide range of tasks, and Retrieval-augmented Generation (RAG) is an effective way to enhance the performance by locating relevant information and placing it into the context window of the LLM. However, the relationship between retrievers and LLMs in a RAG is still under-investigated. Most existing work treats the retriever and the LLM as independent components and leaves a gap between retrieving human-"friendly" information and assembling a LLM-"friendly" context. In this work, we examine a novel bridge mechanism. We validate the ranking and selection assumptions of retrievers in the context of RAG and propose a framework that chains together supervised and reinforcement learning to train a bridge model that optimizes the connection between the retriever and the LLM. Empirical results demonstrate the effectiveness of our method in both question-answering and personalized generation tasks.

cs.CL↗

arXiv2024

On the existence, rareness and uniqueness of quenched HI-rich galaxies in the local Universe

Using data from ALFALFA, xGASS, HI-MaNGA and the Sloan Digital Sky Survey (SDSS), we identify a sample of 47 "red but HI-rich"(RR) galaxies with $NUV-r > 5$ and unusually high HI-to-stellar mass ratios. We compare the optical properties and local environments between the RR galaxies and a control sample of "red and HI-normal"(RN) galaxies that are matched in stellar mass and color. The two samples are similar in the optical properties typical of massive red (quenched) galaxies in the local Universe. The RR sample tends to be associated with slightly lower-density environments and has lower clustering amplitudes and smaller neighbor counts at scales from several kiloparsecs to a few Megaparsecs. The results are consistent with the RR galaxies preferentially being located at the center of low-mass halos, with a median halo mass $\sim 10^{12}h^{-1}M_{\odot}$ compared to $\sim 10^{12.5}h^{-1}M_{\odot}$ for the RN sample. This result is confirmed by the SDSS group catalog which reveals a central fraction of 89% for the RR sample, compared to $\sim 60\%$ for the RN sample. If assumed to follow the HI size-mass relation of normal galaxies, the RR galaxies have an average HI-to-optical radius ratio of $R_{HI}/R_{90}\sim 4$, four times the average ratio for the RN sample. We compare our RR sample with similar samples in previous studies, and quantify the population of RR galaxies using the SDSS complete sample. We conclude that the RR galaxies form a unique but rare population, accounting for only a small fraction of the massive quiescent galaxy population. We discuss the formation scenarios of the RR galaxies.

astro-ph.GA↗

arXiv2024

Learning to Rewrite Prompts for Personalized Text Generation

Facilitated by large language models (LLMs), personalized text generation has become a rapidly growing research direction. Most existing studies focus on designing specialized models for a particular domain, or they require fine-tuning the LLMs to generate personalized text. We consider a typical scenario in which the large language model, which generates personalized output, is frozen and can only be accessed through APIs. Under this constraint, all one can do is to improve the input text (i.e., text prompts) sent to the LLM, a procedure that is usually done manually. In this paper, we propose a novel method to automatically revise prompts for personalized text generation. The proposed method takes the initial prompts generated by a state-of-the-art, multistage framework for personalized generation and rewrites a few critical components that summarize and synthesize the personal context. The prompt rewriter employs a training paradigm that chains together supervised learning (SL) and reinforcement learning (RL), where SL reduces the search space of RL and RL facilitates end-to-end training of the rewriter. Using datasets from three representative domains, we demonstrate that the rewritten prompts outperform both the original prompts and the prompts optimized via supervised learning or reinforcement learning alone. In-depth analysis of the rewritten prompts shows that they are not only human readable, but also able to guide manual revision of prompts when there is limited resource to employ reinforcement learning to train the prompt rewriter, or when it is costly to deploy an automatic prompt rewriter for inference.

cs.CL↗

arXiv2024

Knowledge-driven deep learning for fast MR imaging: undersampled MR image reconstruction from supervised to un-supervised learning

Deep learning (DL) has emerged as a leading approach in accelerating MR imaging. It employs deep neural networks to extract knowledge from available datasets and then applies the trained networks to reconstruct accurate images from limited measurements. Unlike natural image restoration problems, MR imaging involves physics-based imaging processes, unique data properties, and diverse imaging tasks. This domain knowledge needs to be integrated with data-driven approaches. Our review will introduce the significant challenges faced by such knowledge-driven DL approaches in the context of fast MR imaging along with several notable solutions, which include learning neural networks and addressing different imaging application scenarios. The traits and trends of these techniques have also been given which have shifted from supervised learning to semi-supervised learning, and finally, to unsupervised learning methods. In addition, MR vendors' choices of DL reconstruction have been provided along with some discussions on open questions and future directions, which are critical for the reliable imaging systems.

eess.IV↗

arXiv2024

Measuring galaxy abundance and clustering at high redshift from incomplete spectroscopic data: Tests on mock catalogs

The number density and correlation function of galaxies are two key quantities to characterize the distribution of the observed galaxy population. High-$z$ spectroscopic surveys, which usually involve complex target selection and are incomplete in redshift sampling, present both opportunities and challenges to measure these quantities reliably in the high-$z$ Universe. Using realistic mock catalogs we show that target selection and redshift incompleteness can lead to significantly biased results, especially due to the flux limit selection criteria. We develop a new method to correct the flux limit effect, using information provided by the parent photometric data from which the spectroscopic sample is constructed. Our tests using realistic mock samples show that the method is able to reproduce the true stellar mass function and correlation function reliably. Mock catalogs are constructed for the existing zCOSMOS and VIPERS surveys, as well as for the forthcoming PFS galaxy evolution survey. The same set of mock samples are used to quantify the total variance expected for different sample sizes. We find that the total variance decreases very slowly when the survey area reaches about 4 deg$^2$ for the abundance and about 8 deg$^2$ for the clustering, indicating that the cosmic variance is no longer the dominant source of error for PFS-like surveys. We also quantify improvements expected in the PFS-like galaxy survey relative to zCOSMOS and VIPERS surveys.

astro-ph.GA↗

arXiv2024

Post-starburst galaxies in SDSS-IV MaNGA: Two broad categories of evolutionary pathways

We study the size-mass relation (SMR) and recent star formation history (SFH) of post-starburst (PSB) galaxies in the local Universe, using spatially resolved spectroscopy from the final data release of MaNGA. Our sample includes 489 PSB galaxies: 94 cPSB galaxies with central PSB regions, 85 rPSB galaxies with ring-like PSB regions and 310 iPSB galaxies with irregular PSB regions. When compared to control galaxies of similar SFR, redshift and mass, a similar SMR is found for all types of PSB samples except the cPSB galaxies which have smaller sizes at intermediate masses ($9.5\lesssim \log_{10}(\rm M_\ast/M_\odot)\lesssim 10.5$). The iPSB galaxies in the star-forming sequence (iPSB-SF) show no/weak gradients in $\textrm{D}_{n}(4000)$, $\textrm{EW}(\textrm{H}δ_{A})$ and $\textrm{EW}(\textrm{H}α)$, consistent with the global star-forming status of this type of galaxies, while the quiescent iPSB (iPSB-Q) sample shows negative gradients in $\textrm{D}_{n}(4000)$ and positive gradients in $\textrm{EW}(\textrm{H}δ_{A})$, indicating older stellar populations in the inner regions. Both cPSB and rPSB samples show positive gradients in $\textrm{D}_{n}(4000)$ and negative gradients in $\textrm{EW}(\textrm{H}δ_{A})$, indicating younger stellar populations in the inner regions. These results imply that the four types of PSB galaxies can be broadly divided into two distinct categories in terms of evolutionary pathway: (1) iPSB-SF and iPSB-Q which have SMRs and SFHs similar to control galaxies, preferring an inside-out quenching process, (2) rPSB and cPSB which appear to be different stages of the same event, likely to follow the outside-in quenching process driven by disruption events such as mergers that result in a more compact structure as quenching proceeds.

astro-ph.GA↗

arXiv2024

Moist Convective Storms on Saturn

Convective storms (CS) on Saturn appear as bright clouds, irregular in shape and rapidly changing over time, accompanied by lightning phenomena. They cover a wide range of size scales, from 500 km or less and up to sizes that encircle the planet. In this chapter we focus on medium or synoptic-scale events (typical sizes 2,000 - 7,000 km) and planetary-scale major storms known as Great White Spots (GWS). In general, the storms emerge in westward jets and in the peak of the strong eastward equatorial jet. They have been observed preferentially in the northern hemisphere and seem to occur in the same latitude in intervals of 30 and 60 years coinciding with the times of maximum insolation within the seasonal cycle. Moist convection models locate the origin of the storms in the water clouds (at about 10 bars of pressure) and to some extent in ammonia clouds, and their thermodynamics explain the possible periodicity of the phenomena. Once triggered, the storms grow and expand, and their interaction with the zonal wind system gives rise to anticyclonic ovals and to disturbances that propagate zonally covering an entire latitude band in the case of the GWS.

astro-ph.EP↗

arXiv2024

Focus topics for the ECFA study on Higgs / Top / EW factories

In order to stimulate new engagement and trigger some concrete studies in areas where further work would be beneficial towards fully understanding the physics potential of an $e^+e^-$ Higgs / Top / Electroweak factory, we propose to define a set of focus topics. The general reasoning and the proposed topics are described in this document.

hep-ph↗

arXiv2024

DeepSpeed Data Efficiency: Improving Deep Learning Model Quality and Training Efficiency via Efficient Data Sampling and Routing

Recent advances on deep learning models come at the price of formidable training cost. The increasing model size is one of the root causes, but another less-emphasized fact is that data scale is actually increasing at a similar speed as model scale, and the training cost is proportional to both of them. Compared to the rapidly evolving model architecture, how to efficiently use the training data (especially for the expensive foundation model pretraining) is both less explored and difficult to realize due to the lack of a convenient framework that focuses on data efficiency capabilities. To this end, we present DeepSpeed Data Efficiency, a framework that makes better use of data, increases training efficiency, and improves model quality. Specifically, we propose and combine two data efficiency techniques: efficient data sampling via a general curriculum learning library, and efficient data routing via a novel random layerwise token dropping technique. For GPT-3 1.3B language model pretraining, our work achieves 12.5x less data/time/cost (\$3.7K if rent on Azure), while still maintaining 95% of model quality compared to baseline with full data and cost (\$46.3K). For GPT-3 1.3B and BERT-large pretraining, our work can also achieve the same model quality with up to 2x less data/time/cost, or achieve better model quality under same data/time/cost. DeepSpeed Data Efficiency is easy to use and tune, enabling us to easily apply it and verify its benefit on additional tasks including GPT-3 MoE model pretraining and small-scale GPT-2/ViT finetuning.

cs.LG↗

arXiv2024

DarkShot: Lighting Dark Images with Low-Compute and High-Quality

Nighttime photography encounters escalating challenges in extremely low-light conditions, primarily attributable to the ultra-low signal-to-noise ratio. For real-world deployment, a practical solution must not only produce visually appealing results but also require minimal computation. However, most existing methods are either focused on improving restoration performance or employ lightweight models at the cost of quality. This paper proposes a lightweight network that outperforms existing state-of-the-art (SOTA) methods in low-light enhancement tasks while minimizing computation. The proposed network incorporates Siamese Self-Attention Block (SSAB) and Skip-Channel Attention (SCA) modules, which enhance the model's capacity to aggregate global information and are well-suited for high-resolution images. Additionally, based on our analysis of the low-light image restoration process, we propose a Two-Stage Framework that achieves superior results. Our model can restore a UHD 4K resolution image with minimal computation while keeping SOTA restoration quality.

cs.CV↗

arXiv2024

MLIP: Medical Language-Image Pre-training with Masked Local Representation Learning

Existing contrastive language-image pre-training aims to learn a joint representation by matching abundant image-text pairs. However, the number of image-text pairs in medical datasets is usually orders of magnitude smaller than that in natural datasets. Besides, medical image-text pairs often involve numerous complex fine-grained correspondences. This paper aims to enhance the data efficiency by introducing multiple-to-multiple local relationship modeling to capture denser supervisions. More specifically, we propose a Medical Language-Image Pre-training (MLIP) framework, which exploits the limited image-text medical data more efficiently through patch-sentence matching. Furthermore, we introduce a masked contrastive learning strategy with semantic integrity estimation to reduce redundancy in images while preserving the underlying semantics. Our evaluation results show that MLIP outperforms previous work in zero/few-shot classification and few-shot segmentation tasks by a large margin.

cs.CV↗

arXiv2024

LESEN: Label-Efficient deep learning for Multi-parametric MRI-based Visual Pathway Segmentation

Recent research has shown the potential of deep learning in multi-parametric MRI-based visual pathway (VP) segmentation. However, obtaining labeled data for training is laborious and time-consuming. Therefore, it is crucial to develop effective algorithms in situations with limited labeled samples. In this work, we propose a label-efficient deep learning method with self-ensembling (LESEN). LESEN incorporates supervised and unsupervised losses, enabling the student and teacher models to mutually learn from each other, forming a self-ensembling mean teacher framework. Additionally, we introduce a reliable unlabeled sample selection (RUSS) mechanism to further enhance LESEN's effectiveness. Our experiments on the human connectome project (HCP) dataset demonstrate the superior performance of our method when compared to state-of-the-art techniques, advancing multimodal VP segmentation for comprehensive analysis in clinical and research settings. The implementation code will be available at: https://github.com/aldiak/Semi-Supervised-Multimodal-Visual-Pathway- Delineation.

eess.IV↗

arXiv2024

Simultaneous q-Space Sampling Optimization and Reconstruction for Fast and High-fidelity Diffusion Magnetic Resonance Imaging

Diffusion Magnetic Resonance Imaging (dMRI) plays a crucial role in the noninvasive investigation of tissue microstructural properties and structural connectivity in the \textit{in vivo} human brain. However, to effectively capture the intricate characteristics of water diffusion at various directions and scales, it is important to employ comprehensive q-space sampling. Unfortunately, this requirement leads to long scan times, limiting the clinical applicability of dMRI. To address this challenge, we propose SSOR, a Simultaneous q-Space sampling Optimization and Reconstruction framework. We jointly optimize a subset of q-space samples using a continuous representation of spherical harmonic functions and a reconstruction network. Additionally, we integrate the unique properties of diffusion magnetic resonance imaging (dMRI) in both the q-space and image domains by applying $l1$-norm and total-variation regularization. The experiments conducted on HCP data demonstrate that SSOR has promising strengths both quantitatively and qualitatively and exhibits robustness to noise.

cs.CV↗

arXiv2024

Modality Exchange Network for Retinogeniculate Visual Pathway Segmentation

Accurate segmentation of the retinogeniculate visual pathway (RGVP) aids in the diagnosis and treatment of visual disorders by identifying disruptions or abnormalities within the pathway. However, the complex anatomical structure and connectivity of RGVP make it challenging to achieve accurate segmentation. In this study, we propose a novel Modality Exchange Network (ME-Net) that effectively utilizes multi-modal magnetic resonance (MR) imaging information to enhance RGVP segmentation. Our ME-Net has two main contributions. Firstly, we introduce an effective multi-modal soft-exchange technique. Specifically, we design a channel and spatially mixed attention module to exchange modality information between T1-weighted and fractional anisotropy MR images. Secondly, we propose a cross-fusion module that further enhances the fusion of information between the two modalities. Experimental results demonstrate that our method outperforms existing state-of-the-art approaches in terms of RGVP segmentation performance.

eess.IV↗

arXiv2024

AID-DTI: Accelerating High-fidelity Diffusion Tensor Imaging with Detail-Preserving Model-based Deep Learning

Deep learning has shown great potential in accelerating diffusion tensor imaging (DTI). Nevertheless, existing methods tend to suffer from Rician noise and detail loss in reconstructing the DTI-derived parametric maps especially when sparsely sampled q-space data are used. This paper proposes a novel method, AID-DTI (Accelerating hIgh fiDelity Diffusion Tensor Imaging), to facilitate fast and accurate DTI with only six measurements. AID-DTI is equipped with a newly designed Singular Value Decomposition (SVD)-based regularizer, which can effectively capture fine details while suppressing noise during network training. Experimental results on Human Connectome Project (HCP) data consistently demonstrate that the proposed method estimates DTI parameter maps with fine-grained details and outperforms three state-of-the-art methods both quantitatively and qualitatively.

cs.CV↗

arXiv2023

Self-distillation Regularized Connectionist Temporal Classification Loss for Text Recognition: A Simple Yet Effective Approach

Text recognition methods are gaining rapid development. Some advanced techniques, e.g., powerful modules, language models, and un- and semi-supervised learning schemes, consecutively push the performance on public benchmarks forward. However, the problem of how to better optimize a text recognition model from the perspective of loss functions is largely overlooked. CTC-based methods, widely used in practice due to their good balance between performance and inference speed, still grapple with accuracy degradation. This is because CTC loss emphasizes the optimization of the entire sequence target while neglecting to learn individual characters. We propose a self-distillation scheme for CTC-based model to address this issue. It incorporates a framewise regularization term in CTC loss to emphasize individual supervision, and leverages the maximizing-a-posteriori of latent alignment to solve the inconsistency problem that arises in distillation between CTC-based models. We refer to the regularized CTC loss as Distillation Connectionist Temporal Classification (DCTC) loss. DCTC loss is module-free, requiring no extra parameters, longer inference lag, or additional training data or phases. Extensive experiments on public benchmarks demonstrate that DCTC can boost text recognition model accuracy by up to 2.6%, without any of these drawbacks.

cs.CV↗