Search

Model-based reinforcement learning (RL) offers a solution to the data inefficiency that plagues most model-free RL algorithms. However, learning a robust world model often demands complex and deep architectures, which are expensive to compute and train. Within the world model, dynamics models are particularly crucial for accurate predictions, and various dynamics-model architectures have been explored, each with its own set of challenges. Currently, recurrent neural network (RNN) based world models face issues such as vanishing gradients and difficulty in capturing long-term dependencies effectively. In contrast, use of transformers suffers from the well-known issues of self-attention mechanisms, where both memory and computational complexity scale as $O(n^2)$, with $n$ representing the sequence length. To address these challenges we propose a state space model (SSM) based world model, specifically based on Mamba, that achieves $O(n)$ memory and computational complexity while effectively capturing long-term dependencies and facilitating the use of longer training sequences efficiently. We also introduce a novel sampling method to mitigate the suboptimality caused by an incorrect world model in the early stages of training, combining it with the aforementioned technique to achieve a normalised score comparable to other state-of-the-art model-based RL algorithms using only a 7 million trainable parameter world model. This model is accessible and can be trained on an off-the-shelf laptop. Our code is available at https://github.com/realwenlongwang/drama.git.

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.

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.

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., Comment: 11 pages, 7 figures, 1 table

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.

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 water transitions from different upper level energies ($E_u$) and identify single (un-blended) lines that provide the most reliable measurements. With the atlas, we demonstrate two important excitation effects: one related to the opacity saturation of ortho-para line pairs that overlap, and the other to the sub-thermal excitation of $v=1-1$ lines scattered across the $v=0-0$ rotational band. Second, from this larger line selection we define a list of fundamental lines spanning $E_u$ from 1500 to 6000 K to develop simple line-ratio diagrams as diagnostics of temperature components and column density. Third, we report the detection of disk-rotation Doppler broadening of molecular lines, which demonstrates the radial distribution of water emission at different $E_u$ and confirms from gas kinematics a radially-extended $\approx 170$--220 K reservoir, close to the ice sublimation front. We also report the detection of narrow blue-shifted absorption from an inner disk wind in ro-vibrational H2O and CO lines, which may be observed in disks at inclinations $> 50$ deg. We summarize these findings and tools into a general recipe that should be beneficial to community efforts to study water in planet-forming regions., Comment: Community input and feedback is very welcome. Referee report received and asking for minor revisions. Some analysis updates are in progress. The rest of the sample in Appendix H will be included later

Target speaker extraction (TSE) focuses on isolating the speech of a specific target speaker from overlapped multi-talker speech, which is a typical setup in the cocktail party problem. In recent years, TSE draws increasing attention due to its potential for various applications such as user-customized interfaces and hearing aids, or as a crutial front-end processing technologies for subsequential tasks such as speech recognition and speaker recongtion. However, there are currently few open-source toolkits or available pre-trained models for off-the-shelf usage. In this work, we introduce WeSep, a toolkit designed for research and practical applications in TSE. WeSep is featured with flexible target speaker modeling, scalable data management, effective on-the-fly data simulation, structured recipes and deployment support. The toolkit is publicly avaliable at \url{https://github.com/wenet-e2e/WeSep.}, Comment: Interspeech 2024

Deep learning technologies have significantly advanced the performance of target speaker extraction (TSE) tasks. To enhance the generalization and robustness of these algorithms when training data is insufficient, data augmentation is a commonly adopted technique. Unlike typical data augmentation applied to speech mixtures, this work thoroughly investigates the effectiveness of augmenting the enrollment speech space. We found that for both pretrained and jointly optimized speaker encoders, directly augmenting the enrollment speech leads to consistent performance improvement. In addition to conventional methods such as noise and reverberation addition, we propose a novel augmentation method called self-estimated speech augmentation (SSA). Experimental results on the Libri2Mix test set show that our proposed method can achieve an improvement of up to 2.5 dB., Comment: Accepted by SLT2024

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.

FU Ori and EX Lup type objects present natural experiments for understanding a critical stage in the star and planet formation process. These objects offer insight into the diversity of molecules available to forming planetary systems due to a sudden increase in accretion and central luminosity causes the disk and surrounding material to increase in temperature. This allows for volatiles to sublimate off of grains and exist in the gas-phase for tens to hundreds of years post initial outburst. While this dynamic stage may be common for solar-type protostars, observations of the chemical impact of these bursts are rare. In this article, we present observations from the NOrthern Extended Millimeter Array (NOEMA) of five Young Stellar Objects (YSOs) that have undergone outbursts within the past 100 years and catalog the volatile chemistry found within approx 1000 au of the YSO. Only one source clearly shows a line rich spectra with >11 molecules detected including complex organics and water, as is an expected spectra signature for a post-outburst source. This source is V1057 Cyg, and we present it as the northern analog to the well studied and molecule-rich FU Ori source, V883 Ori. Our conclusions on the chemical inventory of the other four sources in our sample are sensitivity limited, as V1057 Cyg contains the highest disk/envelope gas mass., Comment: 11 pages, Accepted to ApJ

Young stellar objects are thought to commonly undergo sudden accretion events that result in a rise in bolometric luminosity. These outbursts likely coincide with the onset of planet formation, and could impact the formation of planets. The reason behind this dramatic enhancement of accretion is an active area of research, and the mass of the system is a critical parameter. Using Northern Extended Millimeter Array, we survey five outbursting sources (three FU Ori, one EX Or, one 'peculiar' source) with the primary goal of determining the system's mass using an optically thin line of CO. We estimate the mass of a central region for each object that using both continuum emission and C17O J=2-1. The C17O emission likely includes both disk and inner envelope material, thus acts as an upper limit on the disk mass, ranging from 0.33-3.4 Msun for our sources. These derived masses suggest that the inner approx. 1000 au contains enough mass along the line of sight for these sources to be gravitationally unstable., Comment: 9 pages, Accepted to ApJ

This study introduces an artificial neural network (ANN) for image classification task, inspired by the aversive olfactory learning circuits of the nematode Caenorhabditis elegans (C. elegans). Despite the remarkable performance of ANNs in a variety of tasks, they face challenges such as excessive parameterization, high training costs and limited generalization capabilities. C. elegans, with its simple nervous system comprising only 302 neurons, serves as a paradigm in neurobiological research and is capable of complex behaviors including learning. This research identifies key neural circuits associated with aversive olfactory learning in C. elegans through behavioral experiments and high-throughput gene sequencing, translating them into an image classification ANN architecture. Additionally, two other image classification ANNs with distinct architectures were constructed for comparative performance analysis to highlight the advantages of bio-inspired design. The results indicate that the ANN inspired by the aversive olfactory learning circuits of C. elegans achieves higher accuracy, better consistency and faster convergence rates in image classification task, especially when tackling more complex classification challenges. This study not only showcases the potential of bio-inspired design in enhancing ANN capabilities but also provides a novel perspective and methodology for future ANN design.

Detecting primordial fluctuations from the cosmic dark ages requires extremely large low-frequency radio telescope arrays deployed on the far side of the Moon. The antenna of such an array must be lightweight, easily storable and transportable, deployable on a large scale, durable, and capable of good electrical performance. A membrane antenna is an excellent candidate to meet these criteria. We study the design of a low-frequency membrane antenna for a lunar-based low-frequency (<30 MHz) radio telescope constructed from polyimide film widely used in aerospace applications, owing to its excellent dielectric properties and high stability as a substrate material. We first design and optimize an antenna in free space through dipole deformation and coupling principles, then simulate an antenna on the lunar surface with a simple lunar soil model, yielding an efficiency greater than 90% in the range of 12-19 MHz and greater than 10% in the range of 5-35 MHz. The antenna inherits the omni-directional radiation pattern of a simple dipole antenna in the 5-30 MHz frequency band, giving a large field of view and allowing detection of the 21 cm global signal when used alone. A demonstration prototype is constructed, and its measured electrical property is found to be consistent with simulated results using |S11| measurements. This membrane antenna can potentially fulfill the requirements of a lunar low-frequency array, establishing a solid technical foundation for future large-scale arrays for exploring the cosmic dark ages., Comment: 14 pages, 19 figures

With the increasing adoption of autonomous vehicles, ensuring the reliability of autonomous driving systems (ADSs) deployed on autonomous vehicles has become a significant concern. Driving simulators have emerged as crucial platforms for testing autonomous driving systems, offering realistic, dynamic, and configurable environments. However, existing simulation-based ADS testers have largely overlooked the reliability of the simulators, potentially leading to overlooked violation scenarios and subsequent safety security risks during real-world deployment. In our investigations, we identified that collision detectors in simulators could fail to detect and report collisions in certain collision scenarios, referred to as ignored collision scenarios. This paper aims to systematically discover ignored collision scenarios to improve the reliability of autonomous driving simulators. To this end, we present ICSFuzz, a black-box fuzzing approach to discover ignored collision scenarios efficiently. Drawing upon the fact that the ignored collision scenarios are a sub-type of collision scenarios, our approach starts with the determined collision scenarios. Following the guidance provided by empirically studied factors contributing to collisions, we selectively mutate arbitrary collision scenarios in a step-wise manner toward the ignored collision scenarios and effectively discover them. We compare ICSFuzz with DriveFuzz, a state-of-the-art simulation-based ADS testing method, by replacing its oracle with our ignored-collision-aware oracle. The evaluation demonstrates that ICSFuzz outperforms DriveFuzz by finding 10-20x more ignored collision scenarios with a 20-70x speedup. All the discovered ignored collisions have been confirmed by developers with one CVE ID assigned.

Bismuth oxyselenide (Bi$_2$O$_2$Se), a novel quasi-2D charge-carrying semiconductor, is hailed as one of the best emerging platforms for the next generation semiconductor devices. Recent efforts on developing diverse Bi$_2$O$_2$Se heterojunctions have produced extensive potential applications in electronics and optoelectronics. In-depth understanding of the band alignment and especially interface dynamics is, however, still challenging. In this work, a comprehensive experimental investigation on the band alignment is performed by a high-resolution X-ray photoelectron spectrometer (HRXPS), and the properties of interface states are also fully discussed. The results show that the ultrathin film Bi$_2$O$_2$Se grown on SrTiO$_3$ (TiO$_2$ (001) termination) exhibits Type-I (straddling gap) band alignment with a valence band offset (VBO) of about 1.77\pm0.04 eV and conduction band offset (CBO) of about 0.68\pm0.04 eV. However, further considering the contribution of the interface states, the bands on the interface present a herringbone configuration due to sizable build-in electric fields, which is significantly different from the conventional band alignment. In this sense, our results provide an insightful guidance to the development of high-efficiency electronic and optoelectronic devices, specifically of the devices where the charge transfer is highly sensitive to interface states.

Optical frequency combs play a crucial role in optical communications, time-frequency metrology, precise ranging, and sensing. Among various generation schemes, resonant electro-optic combs are particularly attractive for its excellent stability, flexibility and broad bandwidths. In this approach, an optical pump undergoes multiple electro-optic modulation processes in a high-Q optical resonator, resulting in cascaded spectral sidebands. However, most resonant electro-optic combs to date make use of lumped-capacitor electrodes with relatively inefficient utilization of the input electrical power. This design also reflects most electrical power back to the driving circuits and necessitates costly RF isolators in between, presenting substantial challenges in practical applications. To address these issues, we present an RF circuit friendly electro-optic frequency comb generator incorporated with on-chip coplanar microwave resonator electrodes, based on a thin-film lithium niobate platform. Our design achieves more than three times electrical power reduction with minimal reflection at the designed comb repetition rate of ~ 25 GHz. We experimentally demonstrate broadband electro-optic frequency comb generation with a comb span of >85 nm at a moderate electrical driving power of 740 mW (28.7 dBm). Our power-efficient and isolator-free electro-optic comb source could offer compact, low-cost and simple-to-design solution for applications in spectroscopy, high-precise metrology, optical communications.

As a novel and promising 2D material, bismuth oxyselenide (Bi$_2$O$_2$Se) has demonstrated significant potential to overcome existing technical barriers in various electronic device applications, due to its unique physical properties like high symmetry, adjustable electronic structure, ultra-high electron mobility. However, the rapid growth of Bi$_2$O$_2$Se films down to a few atomic layers with precise control remains a significant challenge. In this work, the growth of two-dimensional (2D) Bi$_2$O$_2$Se thin films by the pulsed laser deposition (PLD) method is systematically investigated. By controlling temperature, oxygen pressure, laser energy density and laser emission frequency, we successfully prepare atomically thin and flat Bi$_2$O$_2$Se (001) thin films on the (001) surface of SrTiO3. Importantly, we provide a fundamental and unique perspective toward understanding the growth process of atomically thin and flat Bi$_2$O$_2$Se films, and the growth process can be primarily summarized into four steps: i) anisotropic non-spontaneous nucleation preferentially along the step roots; ii) monolayer Bi$_2$O$_2$Se nanosheets expanding across the surrounding area, and eventually covering the entire STO substrate step; iii) vertical growth of Bi$_2$O$_2$Se monolayer in a 2D Frank-van der Merwe (FM) epitaxial growth, and iv) with a layer-by-layer 2D FM growth mode, ultimately producing an atomically flat and epitaxially aligned thin film. Moreover, the combined results of the crystallinity quality, surface morphology and the chemical states manifest the successful PLD-growth of high-quality Bi$_2$O$_2$Se films in a controllable and fast mode.

Detecting planet signatures in protoplanetary disks is fundamental to understanding how and where planets form. In this work, we report dust and gas observational hints of planet formation in the disk around 2MASS-J16120668-301027, as part of the ALMA Large Program "AGE-PRO: ALMA survey of Gas Evolution in Protoplanetary disks". The disk was imaged with the Atacama Large Millimeter/submillimeter Array (ALMA) at Band 6 (1.3 mm) in dust continuum emission and four molecular lines: $^{12}$CO(J=2-1), $^{13}$CO(J=2-1), C$^{18}$O(J=2-1), and H$_2$CO(J=3$_{(3,0)}$-2$_{(2,0)}$). Resolved observations of the dust continuum emission (angular resolution of $\sim 150$ mas, 20 au) show a ring-like structure with a peak at $0.57 ^{\prime \prime}$ (75 au), a deep gap with a minimum at 0.24$^{\prime \prime}$ (31 au), an inner disk, a bridge connecting the inner disk and the outer ring, along with a spiral arm structure, and a tentative detection (to $3\sigma$) of a compact emission at the center of the disk gap, with an estimated dust mass of $\sim 2.7-12.9$ Lunar masses. We also detected a kinematic kink (not coincident with any dust substructure) through several $^{12}$CO channel maps (angular resolution $\sim$ 200 mas, 30 au), located at a radius of $\sim 0.875^{\prime \prime}$ (115.6 au). After modeling the $^{12}$CO velocity rotation around the protostar, we identified a tentative rotating-like structure at the kink location with a geometry similar to that of the disk. We discuss potential explanations for the dust and gas substructures observed in the disk, and their potential connection to signatures of planet formation., Comment: 24 pages, 15 figures

Large Language Models (LLMs) are widely used across various domains, processing millions of daily requests. This surge in demand poses significant challenges in optimizing throughput and latency while keeping costs manageable. The Key-Value (KV) cache, a standard method for retaining previous computations, makes LLM inference highly bounded by memory. While batching strategies can enhance performance, they frequently lead to significant memory fragmentation. Even though cutting-edge systems like vLLM mitigate KV cache fragmentation using paged Attention mechanisms, they still suffer from inefficient memory and computational operations due to the tightly coupled page management and computation kernels. This study introduces the vTensor, an innovative tensor structure for LLM inference based on GPU virtual memory management (VMM). vTensor addresses existing limitations by decoupling computation from memory defragmentation and offering dynamic extensibility. Our framework employs a CPU-GPU heterogeneous approach, ensuring efficient, fragmentation-free memory management while accommodating various computation kernels across different LLM architectures. Experimental results indicate that vTensor achieves an average speedup of 1.86x across different models, with up to 2.42x in multi-turn chat scenarios. Additionally, vTensor provides average speedups of 2.12x and 3.15x in kernel evaluation, reaching up to 3.92x and 3.27x compared to SGLang Triton prefix-prefilling kernels and vLLM paged Attention kernel, respectively. Furthermore, it frees approximately 71.25% (57GB) of memory on the NVIDIA A100 GPU compared to vLLM, enabling more memory-intensive workloads., Comment: 16 pages, 12 figures

We perform the first lattice calculation on the semileptonic decay of $J/\psi$ using the (2+1)-flavor Wilson-clover gauge ensembles generated by CLQCD collaboration. Three gauge ensembles with different lattice spacings, from 0.0519 fm to 0.1053 fm, and pion masses, $m_{\pi}\sim$ 300 MeV, are utilized. After a naive continuum extrapolation using three lattice spacings, we obtain $\operatorname{Br}(J/\psi\rightarrow D_s e\nu_e)=1.90(6)(5)_{V_{cs}}\times 10^{-10}$ and $\operatorname{Br}(J/\psi\rightarrow D e\nu_e)=1.21(6)(9)_{V_{cd}}\times 10^{-11}$, where the first errors are statistical, and the second come from the uncertainties of CKM matrix element $V_{cs(d)}$. The ratios of the branching fractions between lepton $\mu$ and $e$ are also calculated as $R_{J/\psi}(D_s)=0.97002(8)$ and $R_{J/\psi}(D)=0.97423(15)$ after performing a continuum limit including only $a^2$ term. The ratios provide necessary theoretical support for the future experimental test of lepton flavor universality., Comment: 18 pages, 9 figures, accepted by PRD, several images have been enlarged for clarity

Multigrid solvers are the standard in modern scientific computing simulations. Domain Decomposition Aggregation-Based Algebraic Multigrid, also known as the DD-$\alpha$AMG solver, is a successful realization of an algebraic multigrid solver for lattice quantum chromodynamics. Its CPU implementation has made it possible to construct, for some particular discretizations, simulations otherwise computationally unfeasible, and furthermore it has motivated the development and improvement of other algebraic multigrid solvers in the area. From an existing version of DD-$\alpha$AMG already partially ported via CUDA to run some finest-level operations of the multigrid solver on Nvidia GPUs, we translate the CUDA code here by using HIP to run on the ORISE supercomputer. We moreover extend the smoothers available in DD-$\alpha$AMG, paying particular attention to Richardson smoothing, which in our numerical experiments has led to a multigrid solver faster than smoothing with GCR and only 10% slower compared to SAP smoothing. Then we port the odd-even-preconditioned versions of GMRES and Richardson via CUDA. Finally, we extend some computationally intensive coarse-grid operations via advanced vectorization.

The digital correlator is one of the most crucial data processing components of a radio telescope array. With the scale of radio interferometeric array growing, many efforts have been devoted to developing a cost-effective and scalable correlator in the field of radio astronomy. In this paper, a 192-input digital correlator with six CASPER ROACH2 boards and seven GPU servers has been deployed as the digital signal processing system for Tianlai cylinder pathfinder located in Hongliuxia observatory. The correlator consists of 192 input signals (96 dual-polarization), 125-MHz bandwidth, and full-Stokes output. The correlator inherits the advantages of the CASPER system, for example, low cost, high performance, modular scalability, and a heterogeneous computing architecture. With a rapidly deployable ROACH2 digital sampling system, a commercially expandable 10 Gigabit switching network system, and a flexible upgradable GPU computing system, the correlator forms a low-cost and easily-upgradable system, poised to support scalable large-scale interferometeric array in the future., Comment: 12 pages, 13 figures, accepted for publication in Frontiers in Astronomy and Space Sciences, section Astronomical Instrumentation

Pixel-level dense labeling is both resource-intensive and time-consuming, whereas weak labels such as scribble present a more feasible alternative to full annotations. However, training segmentation networks with weak supervision from scribbles remains challenging. Inspired by the fact that different segmentation tasks can be correlated with each other, we introduce a new approach to few-scribble supervised segmentation based on model parameter interpolation, termed as ModelMix. Leveraging the prior knowledge that linearly interpolating convolution kernels and bias terms should result in linear interpolations of the corresponding feature vectors, ModelMix constructs virtual models using convex combinations of convolutional parameters from separate encoders. We then regularize the model set to minimize vicinal risk across tasks in both unsupervised and scribble-supervised way. Validated on three open datasets, i.e., ACDC, MSCMRseg, and MyoPS, our few-scribble guided ModelMix significantly surpasses the performance of the state-of-the-art scribble supervised methods., Comment: 10 pages, 3 figures

Large-scale pretrained vision-language models like CLIP have demonstrated remarkable zero-shot image classification capabilities across diverse domains. To enhance CLIP's performance while preserving the zero-shot paradigm, various test-time prompt tuning methods have been introduced to refine class embeddings through unsupervised learning objectives during inference. However, these methods often encounter challenges in selecting appropriate learning rates to prevent collapsed training in the absence of validation data during test-time adaptation. In this study, we propose a novel backpropagation-free algorithm BaFTA for test-time adaptation of vision-language models. Instead of fine-tuning text prompts to refine class embeddings, our approach directly estimates class centroids using online clustering within a projected embedding space that aligns text and visual embeddings. We dynamically aggregate predictions from both estimated and original class embeddings, as well as from distinct augmented views, by assessing the reliability of each prediction using R\'enyi Entropy. Through extensive experiments, we demonstrate that BaFTA consistently outperforms state-of-the-art test-time adaptation methods in both effectiveness and efficiency., Comment: Preprint updated from our earlier manuscript submitted to ICLR 2024 (https://openreview.net/forum?id=KNtcoAM5Gy)

Optical vector analysis (OVA) is an enabling technology for comprehensively characterizing both amplitude and phase responses of optical devices or systems. Conventional OVA technologies are mostly based on discrete optoelectronic components, leading to unsatisfactory system sizes, complexity, and stability. They also encounter challenges in revealing the on-chip characteristics of integrated photonic devices, which are often overwhelmed by the substantial coupling loss and extra spectral response at chip facets. In this work, we demonstrate a miniaturized OVA system for integrated photonics devices based on broadband single sideband (SSB) modulators on a thin-film lithium niobate (LN) platform. The OVA could provide a direct probe of both amplitude and phase responses of photonic devices with kHz-level resolution and tens of terahertz measurement bandwidth. We perform in-situ characterizations of single and coupled microring resonators fabricated on the same chip as the OVA, unfolding their intrinsic loss and coupling states unambiguously. Furthermore, we achieve the direct measurement of collective phase dynamics and density of states of the Bloch modes in a synthetic frequency crystal, by in-situ OVA of a dynamically modulated microring resonator. Our in-situ OVA system provides a compact, high-precision, and broadband solution for characterizing future integrated photonic devices and circuits, with potential applications ranging from optical communications, biosensing, neuromorphic computing, to quantum information processing.

We prove that for a generic sub-Riemannian and reversible sub-Finsler metrics defined on a fixed co-rank 1 distribution, all strictly normal periodic orbits are non-degenerate.

The training, testing, and deployment, of autonomous vehicles requires realistic and efficient simulators. Moreover, because of the high variability between different problems presented in different autonomous systems, these simulators need to be easy to use, and easy to modify. To address these problems we introduce TorchDriveSim and its benchmark extension TorchDriveEnv. TorchDriveEnv is a lightweight reinforcement learning benchmark programmed entirely in Python, which can be modified to test a number of different factors in learned vehicle behavior, including the effect of varying kinematic models, agent types, and traffic control patterns. Most importantly unlike many replay based simulation approaches, TorchDriveEnv is fully integrated with a state of the art behavioral simulation API. This allows users to train and evaluate driving models alongside data driven Non-Playable Characters (NPC) whose initializations and driving behavior are reactive, realistic, and diverse. We illustrate the efficiency and simplicity of TorchDriveEnv by evaluating common reinforcement learning baselines in both training and validation environments. Our experiments show that TorchDriveEnv is easy to use, but difficult to solve.

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.

MTL is a learning paradigm that effectively leverages both task-specific and shared information to address multiple related tasks simultaneously. In contrast to STL, MTL offers a suite of benefits that enhance both the training process and the inference efficiency. MTL's key advantages encompass streamlined model architecture, performance enhancement, and cross-domain generalizability. Over the past twenty years, MTL has become widely recognized as a flexible and effective approach in various fields, including CV, NLP, recommendation systems, disease prognosis and diagnosis, and robotics. This survey provides a comprehensive overview of the evolution of MTL, encompassing the technical aspects of cutting-edge methods from traditional approaches to deep learning and the latest trend of pretrained foundation models. Our survey methodically categorizes MTL techniques into five key areas: regularization, relationship learning, feature propagation, optimization, and pre-training. This categorization not only chronologically outlines the development of MTL but also dives into various specialized strategies within each category. Furthermore, the survey reveals how the MTL evolves from handling a fixed set of tasks to embracing a more flexible approach free from task or modality constraints. It explores the concepts of task-promptable and -agnostic training, along with the capacity for ZSL, which unleashes the untapped potential of this historically coveted learning paradigm. Overall, we hope this survey provides the research community with a comprehensive overview of the advancements in MTL from its inception in 1997 to the present in 2023. We address present challenges and look ahead to future possibilities, shedding light on the opportunities and potential avenues for MTL research in a broad manner. This project is publicly available at https://github.com/junfish/Awesome-Multitask-Learning., Comment: 60 figures, 116 pages, 500+ references

Planets are formed inside disks around young stars. The gas, dust, and ice in these natal disks are the building materials of planets, and therefore their compositions fundamentally shape the final chemical compositions of planets. In this review, we summarize current observations of molecular lines in protoplanetary disks, from near-infrared to millimeter wavelengths. We discuss the basic types of chemical reactions in disks and the current development of chemical modeling. In particular, we highlight the progress made in understanding snowline locations, abundances of main carriers of carbon, oxygen, and nitrogen, and complex organic molecules in disks. Finally, we discuss efforts to trace planet formation history by combining the understanding of disk chemistry and planet formation processes., Comment: Chapter 2 accepted for publication in the Reviews in Mineralogy and Geochemistry (RiMG) Volume 90 on "Exoplanets: Compositions, Mineralogy, and Evolution" edited by Natalie Hinkel, Keith Putirka, and Siyi Xu; 24 pages and 4 figures

Face Anti-Spoofing (FAS) is pivotal in safeguarding facial recognition systems against presentation attacks. While domain generalization (DG) methods have been developed to enhance FAS performance, they predominantly focus on learning domain-invariant features during training, which may not guarantee generalizability to unseen data that differs largely from the source distributions. Our insight is that testing data can serve as a valuable resource to enhance the generalizability beyond mere evaluation for DG FAS. In this paper, we introduce a novel Test-Time Domain Generalization (TTDG) framework for FAS, which leverages the testing data to boost the model's generalizability. Our method, consisting of Test-Time Style Projection (TTSP) and Diverse Style Shifts Simulation (DSSS), effectively projects the unseen data to the seen domain space. In particular, we first introduce the innovative TTSP to project the styles of the arbitrarily unseen samples of the testing distribution to the known source space of the training distributions. We then design the efficient DSSS to synthesize diverse style shifts via learnable style bases with two specifically designed losses in a hyperspherical feature space. Our method eliminates the need for model updates at the test time and can be seamlessly integrated into not only the CNN but also ViT backbones. Comprehensive experiments on widely used cross-domain FAS benchmarks demonstrate our method's state-of-the-art performance and effectiveness., Comment: Accepted to IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024

We present the first detection of 13CCH in a protoplanetary disk (TW Hya). Using observations of C2H we measure CCH/13CCH = 65 +/- 20 in gas with a CO isotopic ratio of 12CO/13CO = 21 +/- 5 (Yoshida et al. 2022a). The TW Hya disk exhibits a gas phase C/O that exceeds unity and C2H is the tracer of this excess carbon. We confirm that the TW Hya gaseous disk exhibits two separate carbon isotopic reservoirs as noted previously (Yoshida et al. 2022a). We explore two theoretical solutions for the development of this dichotomy. One model represents TW Hya today with a protoplanetary disk exposed to a cosmic ray ionization rate that is below interstellar as consistent with current estimates. We find that this model does not have sufficient ionization in cold (T < 40 K) layers to activate carbon isotopic fractionation. The second model investigates a younger TW Hya protostellar disk exposed to an interstellar cosmic ray ionization rate. We find that the younger model has sources of ionization deeper in a colder disk that generates two independent isotopic reservoirs. One reservoir is 12C-enriched carried by methane/hydrocarbon ices and the other is 13C-enriched carried by gaseous CO. The former potentially provides a source of methane/hydrocarbon ices to power the chemistry that generates the anomalously strong C$_2$H emission in this (and other) disk systems in later stages. The latter provides a source of gaseous 13C-rich material to generate isotopic enrichments in forming giant planets as recently detected in the super-Jupiter TYC 8998-760-1 b by Zhang et al. (2021)., Comment: 16 pages, 8 figures, accepted by the Astrophysical Journal

Machine Learning (ML) has become ubiquitous, fueling data-driven applications across various organizations. Contrary to the traditional perception of ML in research, ML workflows can be complex, resource-intensive, and time-consuming. Expanding an ML workflow to encompass a wider range of data infrastructure and data types may lead to larger workloads and increased deployment costs. Currently, numerous workflow engines are available (with over ten being widely recognized). This variety poses a challenge for end-users in terms of mastering different engine APIs. While efforts have primarily focused on optimizing ML Operations (MLOps) for a specific workflow engine, current methods largely overlook workflow optimization across different engines. In this work, we design and implement Couler, a system designed for unified ML workflow optimization in the cloud. Our main insight lies in the ability to generate an ML workflow using natural language (NL) descriptions. We integrate Large Language Models (LLMs) into workflow generation, and provide a unified programming interface for various workflow engines. This approach alleviates the need to understand various workflow engines' APIs. Moreover, Couler enhances workflow computation efficiency by introducing automated caching at multiple stages, enabling large workflow auto-parallelization and automatic hyperparameters tuning. These enhancements minimize redundant computational costs and improve fault tolerance during deep learning workflow training. Couler is extensively deployed in real-world production scenarios at Ant Group, handling approximately 22k workflows daily, and has successfully improved the CPU/Memory utilization by more than 15% and the workflow completion rate by around 17%.

A diffusion interface two-phase magnetohydrodynamic model has been used for matched densities in our previous work [1,2], which may limit the applications of the model. In this work, we derive a thermodynamically consistent diffuse interface model for diffusion interface two-phase magnetohydrodynamic fluids with large density ratios by Onsager's variational principle and conservation law for the first time. The finite element method for spatial discretization and the first order semi-implicit scheme linked with convect splitting method for temporal discretization, is proposed to solve this new model. The mass conservation, unconditionally energy stability and convergence of the scheme can be proved. Then we derive the existence of weak solutions of governing system employing the above properties of the scheme and compactness method. Finally, we show some numerical results to test the effectiveness and well behavior of proposed scheme.

Molecular emission is used to investigate both the physical and chemical properties of protoplanetary disks. Therefore, to accurately derive disk properties, we need a thorough understanding of the behavior of the molecular probes we rely on. Here we investigate how the molecular line emission of N$_2$H$^+$, HCO$^+$, HCN, and C$^{18}$O compare to other measured quantities in a set of 20 protoplanetary disks. Overall, we find positive correlations between multiple line fluxes and the disk dust mass and radius. We also generally find strong positive correlations between the line fluxes of different molecular species. However, some disks do show noticeable differences in the relative fluxes of N$_2$H$^+$, HCO$^+$, HCN, and C$^{18}$O. These differences occur even within a single star-forming region. This results in a potentially large range of different disk masses and chemical compositions for systems of similar age and birth environment. While we make preliminary comparisons of molecular fluxes across different star-forming regions, more complete and uniform samples are needed in the future to search for trends with birth environment or age., Comment: Accepted for publication in The Astrophysical Journal

Training for multi-agent reinforcement learning(MARL) is a time-consuming process caused by distribution shift of each agent. One drawback is that strategy of each agent in MARL is independent but actually in cooperation. Thus, a vertical issue in multi-agent reinforcement learning is how to efficiently accelerate training process. To address this problem, current research has leveraged a centralized function(CF) across multiple agents to learn contribution of the team reward for each agent. However, CF based methods introduce joint error from other agents in estimation of value network. In so doing, inspired by federated learning, we propose three simple novel approaches called Average Periodically Parameter Sharing(A-PPS), Reward-Scalability Periodically Parameter Sharing(RS-PPS) and Partial Personalized Periodically Parameter Sharing(PP-PPS) mechanism to accelerate training of MARL. Agents share Q-value network periodically during the training process. Agents which has same identity adapt collected reward as scalability and update partial neural network during period to share different parameters. We apply our approaches in classical MARL method QMIX and evaluate our approaches on various tasks in StarCraft Multi-Agent Challenge(SMAC) environment. Performance of numerical experiments yield enormous enhancement, with an average improvement of 10\%-30\%, and enable to win tasks that QMIX cannot. Our code can be downloaded from https://github.com/ColaZhang22/PPS-QMIX, Comment: 10 pages, 5 figures

Score function estimation is the cornerstone of both training and sampling from diffusion generative models. Despite this fact, the most commonly used estimators are either biased neural network approximations or high variance Monte Carlo estimators based on the conditional score. We introduce a novel nearest neighbour score function estimator which utilizes multiple samples from the training set to dramatically decrease estimator variance. We leverage our low variance estimator in two compelling applications. Training consistency models with our estimator, we report a significant increase in both convergence speed and sample quality. In diffusion models, we show that our estimator can replace a learned network for probability-flow ODE integration, opening promising new avenues of future research., Comment: 25 pages, 9 figures. To be published in ICML 2024

Real-time semantic segmentation is a crucial research for real-world applications. However, many methods lay particular emphasis on reducing the computational complexity and model size, while largely sacrificing the accuracy. To tackle this problem, we propose a parallel inference network customized for semantic segmentation tasks to achieve a good trade-off between speed and accuracy. We employ a shallow backbone to ensure real-time speed, and propose three core components to compensate for the reduced model capacity to improve accuracy. Specifically, we first design a dual-pyramidal path architecture (Multi-level Feature Aggregation Module, MFAM) to aggregate multi-level features from the encoder to each scale, providing hierarchical clues for subsequent spatial alignment and corresponding in-network inference. Then, we build Recursive Alignment Module (RAM) by combining the flow-based alignment module with recursive upsampling architecture for accurate spatial alignment between multi-scale feature maps with half the computational complexity of the straightforward alignment method. Finally, we perform independent parallel inference on the aligned features to obtain multi-scale scores, and adaptively fuse them through an attention-based Adaptive Scores Fusion Module (ASFM) so that the final prediction can favor objects of multiple scales. Our framework shows a better balance between speed and accuracy than state-of-the-art real-time methods on Cityscapes and CamVid datasets. We also conducted systematic ablation studies to gain insight into our motivation and architectural design. Code is available at: https://github.com/Yanhua-Zhang/MFARANet., Comment: 15 pages, 9 figures and 12 Tables. Manuscript completed on April 30, 2022