arXiv subjects

Ke Zhang

Publications and source records attributed to Ke Zhang.

At least 163 records · Page 9Linked to original sources

arXiv2025

Smuggling unnoticed: Towards a 2D view of water and dust delivery to the inner regions of protoplanetary discs

Infrared spectroscopy, e.g., with JWST, provides a glimpse into the chemical inventory of the innermost region of protoplanetary discs, where terrestrial planets eventually form. The chemical make-up of regions inside snowlines is connected to the material drifting from the outer regions, which can be modeled with dust evolution models. However, infrared observations are limited by the high dust extinction in the inner disc, and only probes the abundances of gaseous species in the disc surface layers. As a result, the bulk mass of delivered volatiles is not directly relatable to what is measured through infrared spectra. In this paper, we investigate how the delivery of dust and ice after prolonged pebble drift affects the observable reservoir of water vapor in the inner disc. We develop a 1+1D approach based on dust evolution models to determine the delivery and distribution of vapor compared to the height of the $τ= 1$ surface in the dust continuum. We find that the observable column density of water vapor at wavelengths probed by JWST spans many orders of magnitude over time, exhibiting different radial profiles depending on dust properties, drift rate, and local processing. In the presence of a traffic-jam effect inside the snowline, the observable vapor reservoir appears constant in time despite the ongoing delivery by pebble drift, such that water is effectively smuggled unnoticed. Differences in measured column densities then originate not only from variations in bulk vapor content, but also from differences in the properties and distribution of dust particles.

astro-ph.EP↗

arXiv2024

The First JWST View of a 30-Myr-old Protoplanetary Disk Reveals a Late-stage Carbon-rich Phase

We present a JWST MIRI/MRS spectrum of the inner disk of WISE J044634.16$-$262756.1B (hereafter J0446B), an old ($\sim$34 Myr) M4.5 star but with hints of ongoing accretion. The spectrum is molecule-rich and dominated by hydrocarbons. We detect 14 molecular species (H$_2$, CH$_3$, CH$_4$, C$_2$H$_2$, $^{13}$CCH$_2$, C$_2$H$_4$, C$_2$H$_6$, C$_3$H$_4$, C$_4$H$_2$, C$_6$H$_6$, HCN, HC$_3$N, CO$_2$ and $^{13}$CO$_2$) and 2 atomic lines ([Ne II] and [Ar II]), all observed for the first time in a disk at this age. The detection of spatially unresolved H$_2$ and Ne gas strongly supports that J0446B hosts a long-lived primordial disk, rather than a debris disk. The marginal H$_2$O detection and the high C$_2$H$_2$/CO$_2$ column density ratio indicate that the inner disk of J0446B has a very carbon-rich chemistry, with a gas-phase C/O ratio $\gtrsim$2, consistent with what have been found in most primordial disks around similarly low-mass stars. In the absence of significant outer disk dust substructures, inner disks are expected to first become water-rich due to the rapid inward drift of icy pebbles, and evolve into carbon-rich as outer disk gas flows inward on longer timescales. The faint millimeter emission in such low-mass star disks implies that they may have depleted their outer icy pebble reservoir early and already passed the water-rich phase. Models with pebble drift and volatile transport suggest that maintaining a carbon-rich chemistry for tens of Myr likely requires a slowly evolving disk with $α-$viscosity $\lesssim10^{-4}$. This study represents the first detailed characterization of disk gas at $\sim$30 Myr, strongly motivating further studies into the final stages of disk evolution.

astro-ph.EP↗

arXiv2024

V-MIND: Building Versatile Monocular Indoor 3D Detector with Diverse 2D Annotations

The field of indoor monocular 3D object detection is gaining significant attention, fueled by the increasing demand in VR/AR and robotic applications. However, its advancement is impeded by the limited availability and diversity of 3D training data, owing to the labor-intensive nature of 3D data collection and annotation processes. In this paper, we present V-MIND (Versatile Monocular INdoor Detector), which enhances the performance of indoor 3D detectors across a diverse set of object classes by harnessing publicly available large-scale 2D datasets. By leveraging well-established monocular depth estimation techniques and camera intrinsic predictors, we can generate 3D training data by converting large-scale 2D images into 3D point clouds and subsequently deriving pseudo 3D bounding boxes. To mitigate distance errors inherent in the converted point clouds, we introduce a novel 3D self-calibration loss for refining the pseudo 3D bounding boxes during training. Additionally, we propose a novel ambiguity loss to address the ambiguity that arises when introducing new classes from 2D datasets. Finally, through joint training with existing 3D datasets and pseudo 3D bounding boxes derived from 2D datasets, V-MIND achieves state-of-the-art object detection performance across a wide range of classes on the Omni3D indoor dataset.

cs.CV↗

arXiv2024

Generic global diffusion for analytic uncoupled a priori unstable systems

We show that given a general uncoupled a priori unstable Hamiltonian \[ \frac12 p^2 + V(q) + G(I) + εh(p, q, I, φ, t), \] where $h$ is a generic Mañé analytic function and $ε$ is small enough, there is an orbit for which the momentum $I$ changes by any arbitrarily prescribed value. We call this phenomenon as global diffusion since the size of the change in $I$ is independent of both $ε$ and $h$. The fact that the pendulum and rotor variables are uncoupled is used essentially in our proof. The proof is based on simple and constructive geometrical methods, carefully studying the reduced Poincaré functions of the problem which generate the corresponding scattering maps.

math.DS↗

arXiv2024

Multi-Level Speaker Representation for Target Speaker Extraction

Target speaker extraction (TSE) relies on a reference cue of the target to extract the target speech from a speech mixture. While a speaker embedding is commonly used as the reference cue, such embedding pre-trained with a large number of speakers may suffer from confusion of speaker identity. In this work, we propose a multi-level speaker representation approach, from raw features to neural embeddings, to serve as the speaker reference cue. We generate a spectral-level representation from the enrollment magnitude spectrogram as a raw, low-level feature, which significantly improves the model's generalization capability. Additionally, we propose a contextual embedding feature based on cross-attention mechanisms that integrate frame-level embeddings from a pre-trained speaker encoder. By incorporating speaker features across multiple levels, we significantly enhance the performance of the TSE model. Our approach achieves a 2.74 dB improvement and a 4.94% increase in extraction accuracy on Libri2mix test set over the baseline.

eess.AS↗

arXiv2024

AGD: an Auto-switchable Optimizer using Stepwise Gradient Difference for Preconditioning Matrix

Adaptive optimizers, such as Adam, have achieved remarkable success in deep learning. A key component of these optimizers is the so-called preconditioning matrix, providing enhanced gradient information and regulating the step size of each gradient direction. In this paper, we propose a novel approach to designing the preconditioning matrix by utilizing the gradient difference between two successive steps as the diagonal elements. These diagonal elements are closely related to the Hessian and can be perceived as an approximation of the inner product between the Hessian row vectors and difference of the adjacent parameter vectors. Additionally, we introduce an auto-switching function that enables the preconditioning matrix to switch dynamically between Stochastic Gradient Descent (SGD) and the adaptive optimizer. Based on these two techniques, we develop a new optimizer named AGD that enhances the generalization performance. We evaluate AGD on public datasets of Natural Language Processing (NLP), Computer Vision (CV), and Recommendation Systems (RecSys). Our experimental results demonstrate that AGD outperforms the state-of-the-art (SOTA) optimizers, achieving highly competitive or significantly better predictive performance. Furthermore, we analyze how AGD is able to switch automatically between SGD and the adaptive optimizer and its actual effects on various scenarios. The code is available at https://github.com/intelligent-machine-learning/atorch/tree/main/atorch/optimizers.

cs.LG↗

arXiv2024

Water in protoplanetary disks with JWST-MIRI: spectral excitation atlas and radial distribution from temperature diagnostic diagrams and Doppler mapping

This work aims at providing fundamental general tools for the analysis of water spectra as observed in protoplanetary disks with JWST-MIRI. We analyze 25 high-quality spectra from the JDISC Survey reduced with asteroid calibrators as presented in Pontoppidan et al. (2024). First, we present a spectral atlas to illustrate the clustering of H$_2$O transitions from different upper level energies ($E_u$) and identify single (un-blended) transitions that provide the most reliable measurements. With that, we demonstrate two important excitation effects: the opacity saturation of ortho-para line pairs that overlap, and the non-LTE excitation of $v=1-1$ lines scattered across the $v=0-0$ rotational band. Second, we define a shorter list of fundamental lines spanning $E_u=$ 1500-6000 K to develop simple line-ratio diagnostic diagrams for the radial temperature distribution of water in inner disks, which can be interpreted using discrete temperature components or a radial gradient. Third, we report the detection of disk-rotation Doppler broadening of molecular lines, which confirms the radial distribution of water emission including, for the first time, the radially-extended $\approx$ 170-220 K reservoir close to the snowline. The combination of measured line ratios and broadening suggests that drift-dominated disks have shallower temperature gradients with an extended cooler disk surface enriched by ice sublimation. We also report the first detection of a H$_2$O-rich inner disk wind from narrow blue-shifted absorption in the ro-vibrational lines. We summarize these findings and tools into a general recipe to make the study of water in planet-forming regions reliable, effective, and sustainable for samples of $> 100$ disks.

astro-ph.EP↗

arXiv2024

Sharpness-Aware Minimization Revisited: Weighted Sharpness as a Regularization Term

Deep Neural Networks (DNNs) generalization is known to be closely related to the flatness of minima, leading to the development of Sharpness-Aware Minimization (SAM) for seeking flatter minima and better generalization. In this paper, we revisit the loss of SAM and propose a more general method, called WSAM, by incorporating sharpness as a regularization term. We prove its generalization bound through the combination of PAC and Bayes-PAC techniques, and evaluate its performance on various public datasets. The results demonstrate that WSAM achieves improved generalization, or is at least highly competitive, compared to the vanilla optimizer, SAM and its variants. The code is available at https://github.com/intelligent-machine-learning/atorch/tree/main/atorch/optimizers.

cs.LG↗

arXiv2024

Mono-drive single-sideband modulation via optical delay lines on thin-film lithium niobate

Optical single-sideband (SSB) modulation features high spectral efficiency, substantial dispersion tolerance, and straightforward detection, making it a versatile technology for applications in optical communications, microwave photonics, optical sensing, satellite communication, etc. However, conventional SSB generators typically require two radio-frequency (RF) signals with a 90° phase difference to drive a pair of parallel phase or amplitude modulators, resulting in high system complexity and low power efficiency. In this paper, we propose and realize a simplified SSB generation scheme necessitating only a single RF drive, by achieving effective RF phase shift using on-chip optical delay lines. This approach not only reduces system complexity and saves energy consumption by 3 dB, but also enables easy scalability to higher frequencies. We demonstrate both full-carrier SSB (FC-SSB) and carrier-suppressed SSB (CS-SSB) modulation on thin-film lithium niobate platform. For FC-SSB, we show a maximum sideband suppression of 22.1 dB at 50 GHz and apply it to address the frequency-selective power fading problem in optical communication systems. For CS-SSB, we show a maximum sideband suppression of 22.5 dB and a sideband-to-carrier suppression of 16.9 dB at 50 GHz, which can act as an optical frequency shifter by sweeping the modulation frequencies. Moreover, the shifted optical frequency can be transferred back to the electrical domain by beating with a reference signal generated via a phase modulator on the same chip, achieving broadband RF frequency shifting from a maximum of 50 GHz down to 1 GHz. Our simple, power-efficient, and low-cost SSB modulation scheme could provide an effective solution for future high-frequency direct detection-based communication systems, frequency-modulated continuous wave radar/LiDAR, optical vector network analyzers, and microwave photonics systems.

physics.optics↗

arXiv2024

MA^2: A Self-Supervised and Motion Augmenting Autoencoder for Gait-Based Automatic Disease Detection

Ground reaction force (GRF) is the force exerted by the ground on a body in contact with it. GRF-based automatic disease detection (ADD) has become an emerging medical diagnosis method, which aims to learn and identify disease patterns corresponding to different gait pressures based on deep learning methods. Although existing ADD methods can save doctors time in making diagnoses, training deep models still struggles with the cost caused by the labeling engineering for a large number of gait diagnostic data for subjects. On the other hand, the accuracy of the deep model under the unified benchmark GRF dataset and the generalization ability on scalable gait datasets need to be further improved. To address these issues, we propose MA2, a GRF-based self-supervised and motion augmenting auto-encoder, which models the ADD task as an encoder-decoder paradigm. In the encoder, we introduce an embedding block including the 3-layer 1D convolution for extracting the token and a mask generator to randomly mask out the sequence of tokens to maximize the model's potential to capture high-level, discriminative, intrinsic representations. whereafter, the decoder utilizes this information to reconstruct the pixel sequence of the origin input and calculate the reconstruction loss to optimize the network. Moreover, the backbone of an auto-encoder is multi-head self-attention that can consider the global information of the token from the input, not just the local neighborhood. This allows the model to capture generalized contextual information. Extensive experiments demonstrate MA2 has SOTA performance of 90.91% accuracy on 1% limited pathological GRF samples with labels, and good generalization ability of 78.57% accuracy on scalable Parkinson disease dataset.

physics.bio-ph↗

arXiv2024

DFabric: Scaling Out Data Parallel Applications with CXL-Ethernet Hybrid Interconnects

Emerging interconnects, such as CXL and NVLink, have been integrated into the intra-host topology to scale more accelerators and facilitate efficient communication between them, such as GPUs. To keep pace with the accelerator's growing computing throughput, the interconnect has seen substantial enhancement in link bandwidth, e.g., 256GBps for CXL 3.0 links, which surpasses Ethernet and InfiniBand network links by an order of magnitude or more. Consequently, when data-intensive jobs, such as LLM training, scale across multiple hosts beyond the reach limit of the interconnect, the performance is significantly hindered by the limiting bandwidth of the network infrastructure. We address the problem by proposing DFabric, a two-tier interconnect architecture. We address the problem by proposing DFabric, a two-tier interconnect architecture. First, DFabric disaggregates rack's computing units with an interconnect fabric, i.e., CXL fabric, which scales at rack-level, so that they can enjoy intra-rack efficient interconnecting. Second, DFabric disaggregates NICs from hosts, and consolidates them to form a NIC pool with CXL fabric. By providing sufficient aggregated capacity comparable to interconnect bandwidth, the NIC pool bridges efficient communication across racks or beyond the reach limit of interconnect fabric. However, the local memory accessing becomes the bottleneck when enabling each host to utilize the NIC pool efficiently. To the end, DFabric builds a memory pool with sufficient bandwidth by disaggregating host local memory and adding more memory devices. We have implemented a prototype of DFabric that can run applications transparently. We validated its performance gain by running various microbenchmarks and compute-intensive applications such as DNN and graph.

cs.DC↗

arXiv2024

Hypermultiplexed Integrated-Photonics-based Tensor Optical Processor

The escalating data volume and complexity resulting from the rapid expansion of artificial intelligence (AI), internet of things (IoT) and 5G/6G mobile networks is creating an urgent need for energy-efficient, scalable computing hardware. Here we demonstrate a hypermultiplexed integratedphotonics-based tensor optical processor (HITOP) that can perform trillions of operations per second (TOPS) at the energy efficiency of 40 TOPS/W. Space-time-wavelength three-dimensional (3D) optical parallelism enables O($N^{2}$) operations per clock-cycle using O($N$) modulator devices. The system is built with wafer-fabricated III/V micron-scale lasers and high-speed thin-film Lithium-Niobate electro-optics for encoding at 10s femtojoule/symbol. Lasing threshold incorporates analog inline rectifier (ReLu) nonlinearity for low-latency activation. The system scalability is verified with machine learning models of 405,000 parameters. A combination of high clockrates, energy-efficient processing and programmability unlocks the potential of light for large-scale AI accelerators in applications ranging from training of large AI models to real-time decision making in edge deployment.

cs.ET↗

arXiv2024

Broadband millimeter-wave frequency mixer based on thin-film lithium niobate photonics

Frequency mixers are fundamental components in modern wireless communication and radar systems, responsible for up- and down-conversion of target radio-frequency (RF) signals. Recently, photonic-assisted RF mixers have shown unique advantages over traditional electronic counterparts, including broad operational bandwidth, flat frequency response, and immunity to electromagnetic interference. However, current integrated photonic mixers face significant challenges in achieving efficient conversion at high frequencies, especially in millimeter-wave bands, due to the limitations of existing electro-optic (EO) modulators. Additionally, high-frequency local oscillators in the millimeter-wave range are often difficult to obtain and expensive, leading to unsatisfactory cost and restricted operational bandwidth in practice. In this paper, we harness the exceptional EO property and scalability of thin-film lithium niobate (TFLN) photonic platform to implement a high-performance harmonic reconfigurable millimeter-wave mixer. The TFLN photonic circuit integrates a broadband EO modulator that allows for extensive frequency coverage, and an EO frequency comb source that significantly reduces the required carrier frequency of the local oscillator. We experimentally demonstrate fully reconfigurable frequency down-conversion across a broad operational bandwidth ranging from 20 GHz to 67 GHz, with a large intermediate frequency of 20 GHz, as well as up-conversion to frequencies of up to 110 GHz. Our integrated photonic mixing system shows dramatically improved bandwidth performance, along with competitive indicators of frequency conversion efficiency and spurious suppression ratio, positioning it as a promising solution for future millimeter-wave transceivers in next-generation communication and sensing systems.

physics.optics↗

arXiv2024

HumVI: A Multilingual Dataset for Detecting Violent Incidents Impacting Humanitarian Aid

Humanitarian organizations can enhance their effectiveness by analyzing data to discover trends, gather aggregated insights, manage their security risks, support decision-making, and inform advocacy and funding proposals. However, data about violent incidents with direct impact and relevance for humanitarian aid operations is not readily available. An automatic data collection and NLP-backed classification framework aligned with humanitarian perspectives can help bridge this gap. In this paper, we present HumVI - a dataset comprising news articles in three languages (English, French, Arabic) containing instances of different types of violent incidents categorized by the humanitarian sector they impact, e.g., aid security, education, food security, health, and protection. Reliable labels were obtained for the dataset by partnering with a data-backed humanitarian organization, Insecurity Insight. We provide multiple benchmarks for the dataset, employing various deep learning architectures and techniques, including data augmentation and mask loss, to address different task-related challenges, e.g., domain expansion. The dataset is publicly available at https://github.com/dataminr-ai/humvi-dataset.

cs.CL↗

arXiv2024

Integrated adaptive coherent LiDAR for 4D bionic vision

Light detection and ranging (LiDAR) is a ubiquitous tool to provide precise spatial awareness in various perception environments. A bionic LiDAR that can mimic human-like vision by adaptively gazing at selected regions of interest within a broad field of view is crucial to achieve high-resolution imaging in an energy-saving and cost-effective manner. However, current LiDARs based on stacking fixed-wavelength laser arrays and inertial scanning have not been able to achieve the desired dynamic focusing patterns and agile scalability simultaneously. Moreover, the ability to synchronously acquire multi-dimensional physical parameters, including distance, direction, Doppler, and color, through seamless fusion between multiple sensors, still remains elusive in LiDAR. Here, we overcome these limitations and demonstrate a bio-inspired frequency-modulated continuous wave (FMCW) LiDAR system with dynamic and scalable gazing capability. Our chip-scale LiDAR system is built using hybrid integrated photonic solutions, where a frequency-chirped external cavity laser provides broad spectral tunability, while on-chip electro-optic combs with elastic channel spacing allow customizable imaging granularity. Using the dynamic zoom-in capability and the coherent FMCW scheme, we achieve a state-of-the-art resolution of 0.012 degrees, providing up to 15 times the resolution of conventional 3D LiDAR sensors, with 115 equivalent scanning lines and 4D parallel imaging. We further demonstrate cooperative sensing between our adaptive coherent LiDAR and a camera to enable high-resolution color-enhanced machine vision.

physics.optics↗

arXiv2024

LayerKV: Optimizing Large Language Model Serving with Layer-wise KV Cache Management

The expanding context windows in large language models (LLMs) have greatly enhanced their capabilities in various applications, but they also introduce significant challenges in maintaining low latency, particularly in Time to First Token (TTFT). This paper identifies that the sharp rise in TTFT as context length increases is predominantly driven by queuing delays, which are caused by the growing demands for GPU Key-Value (KV) cache allocation clashing with the limited availability of KV cache blocks. To address this issue, we propose LayerKV, a simple yet effective plug-in method that effectively reduces TTFT without requiring additional hardware or compromising output performance, while seamlessly integrating with existing parallelism strategies and scheduling techniques. Specifically, LayerKV introduces layer-wise KV block allocation, management, and offloading for fine-grained control over system memory, coupled with an SLO-aware scheduler to optimize overall Service Level Objectives (SLOs). Comprehensive evaluations on representative models, ranging from 7B to 70B parameters, across various GPU configurations, demonstrate that LayerKV improves TTFT latency up to 69x and reduces SLO violation rates by 28.7%, significantly enhancing the user experience.

cs.DC↗

arXiv2024

Semantically Consistent Video Inpainting with Conditional Diffusion Models

Current state-of-the-art methods for video inpainting typically rely on optical flow or attention-based approaches to inpaint masked regions by propagating visual information across frames. While such approaches have led to significant progress on standard benchmarks, they struggle with tasks that require the synthesis of novel content that is not present in other frames. In this paper, we reframe video inpainting as a conditional generative modeling problem and present a framework for solving such problems with conditional video diffusion models. We introduce inpainting-specific sampling schemes which capture crucial long-range dependencies in the context, and devise a novel method for conditioning on the known pixels in incomplete frames. We highlight the advantages of using a generative approach for this task, showing that our method is capable of generating diverse, high-quality inpaintings and synthesizing new content that is spatially, temporally, and semantically consistent with the provided context.

cs.CV↗

arXiv2024

Semi-LLIE: Semi-supervised Contrastive Learning with Mamba-based Low-light Image Enhancement

Despite the impressive advancements made in recent low-light image enhancement techniques, the scarcity of paired data has emerged as a significant obstacle to further advancements. This work proposes a mean-teacher-based semi-supervised low-light enhancement (Semi-LLIE) framework that integrates the unpaired data into model training. The mean-teacher technique is a prominent semi-supervised learning method, successfully adopted for addressing high-level and low-level vision tasks. However, two primary issues hinder the naive mean-teacher method from attaining optimal performance in low-light image enhancement. Firstly, pixel-wise consistency loss is insufficient for transferring realistic illumination distribution from the teacher to the student model, which results in color cast in the enhanced images. Secondly, cutting-edge image enhancement approaches fail to effectively cooperate with the mean-teacher framework to restore detailed information in dark areas due to their tendency to overlook modeling structured information within local regions. To mitigate the above issues, we first introduce a semantic-aware contrastive loss to faithfully transfer the illumination distribution, contributing to enhancing images with natural colors. Then, we design a Mamba-based low-light image enhancement backbone to effectively enhance Mamba's local region pixel relationship representation ability with a multi-scale feature learning scheme, facilitating the generation of images with rich textural details. Further, we propose novel perceptive loss based on the large-scale vision-language Recognize Anything Model (RAM) to help generate enhanced images with richer textual details. The experimental results indicate that our Semi-LLIE surpasses existing methods in both quantitative and qualitative metrics.

cs.CV↗