arXiv subjects

Ming Yan

Publications and source records attributed to Ming Yan.

At least 199 records · Page 11Linked to original sources

arXiv2023

Construction and Applications of Billion-Scale Pre-Trained Multimodal Business Knowledge Graph

Business Knowledge Graphs (KGs) are important to many enterprises today, providing factual knowledge and structured data that steer many products and make them more intelligent. Despite their promising benefits, building business KG necessitates solving prohibitive issues of deficient structure and multiple modalities. In this paper, we advance the understanding of the practical challenges related to building KG in non-trivial real-world systems. We introduce the process of building an open business knowledge graph (OpenBG) derived from a well-known enterprise, Alibaba Group. Specifically, we define a core ontology to cover various abstract products and consumption demands, with fine-grained taxonomy and multimodal facts in deployed applications. OpenBG is an open business KG of unprecedented scale: 2.6 billion triples with more than 88 million entities covering over 1 million core classes/concepts and 2,681 types of relations. We release all the open resources (OpenBG benchmarks) derived from it for the community and report experimental results of KG-centric tasks. We also run up an online competition based on OpenBG benchmarks, and has attracted thousands of teams. We further pre-train OpenBG and apply it to many KG- enhanced downstream tasks in business scenarios, demonstrating the effectiveness of billion-scale multimodal knowledge for e-commerce. All the resources with codes have been released at \url{https://github.com/OpenBGBenchmark/OpenBG}.

cs.AI↗

arXiv2023

Communication-Efficient Topologies for Decentralized Learning with $O(1)$ Consensus Rate

Decentralized optimization is an emerging paradigm in distributed learning in which agents achieve network-wide solutions by peer-to-peer communication without the central server. Since communication tends to be slower than computation, when each agent communicates with only a few neighboring agents per iteration, they can complete iterations faster than with more agents or a central server. However, the total number of iterations to reach a network-wide solution is affected by the speed at which the agents' information is ``mixed'' by communication. We found that popular communication topologies either have large maximum degrees (such as stars and complete graphs) or are ineffective at mixing information (such as rings and grids). To address this problem, we propose a new family of topologies, EquiTopo, which has an (almost) constant degree and a network-size-independent consensus rate that is used to measure the mixing efficiency. In the proposed family, EquiStatic has a degree of $Θ(\ln(n))$, where $n$ is the network size, and a series of time-dependent one-peer topologies, EquiDyn, has a constant degree of 1. We generate EquiDyn through a certain random sampling procedure. Both of them achieve an $n$-independent consensus rate. We apply them to decentralized SGD and decentralized gradient tracking and obtain faster communication and better convergence, theoretically and empirically. Our code is implemented through BlueFog and available at \url{https://github.com/kexinjinnn/EquiTopo}

math.OC↗

arXiv2023

Learning Trajectory-Word Alignments for Video-Language Tasks

In a video, an object usually appears as the trajectory, i.e., it spans over a few spatial but longer temporal patches, that contains abundant spatiotemporal contexts. However, modern Video-Language BERTs (VDL-BERTs) neglect this trajectory characteristic that they usually follow image-language BERTs (IL-BERTs) to deploy the patch-to-word (P2W) attention that may over-exploit trivial spatial contexts and neglect significant temporal contexts. To amend this, we propose a novel TW-BERT to learn Trajectory-Word alignment by a newly designed trajectory-to-word (T2W) attention for solving video-language tasks. Moreover, previous VDL-BERTs usually uniformly sample a few frames into the model while different trajectories have diverse graininess, i.e., some trajectories span longer frames and some span shorter, and using a few frames will lose certain useful temporal contexts. However, simply sampling more frames will also make pre-training infeasible due to the largely increased training burdens. To alleviate the problem, during the fine-tuning stage, we insert a novel Hierarchical Frame-Selector (HFS) module into the video encoder. HFS gradually selects the suitable frames conditioned on the text context for the later cross-modal encoder to learn better trajectory-word alignments. By the proposed T2W attention and HFS, our TW-BERT achieves SOTA performances on text-to-video retrieval tasks, and comparable performances on video question-answering tasks with some VDL-BERTs trained on much more data. The code will be available in the supplementary material.

cs.CV↗

arXiv2023

mPLUG-2: A Modularized Multi-modal Foundation Model Across Text, Image and Video

Recent years have witnessed a big convergence of language, vision, and multi-modal pretraining. In this work, we present mPLUG-2, a new unified paradigm with modularized design for multi-modal pretraining, which can benefit from modality collaboration while addressing the problem of modality entanglement. In contrast to predominant paradigms of solely relying on sequence-to-sequence generation or encoder-based instance discrimination, mPLUG-2 introduces a multi-module composition network by sharing common universal modules for modality collaboration and disentangling different modality modules to deal with modality entanglement. It is flexible to select different modules for different understanding and generation tasks across all modalities including text, image, and video. Empirical study shows that mPLUG-2 achieves state-of-the-art or competitive results on a broad range of over 30 downstream tasks, spanning multi-modal tasks of image-text and video-text understanding and generation, and uni-modal tasks of text-only, image-only, and video-only understanding. Notably, mPLUG-2 shows new state-of-the-art results of 48.0 top-1 accuracy and 80.3 CIDEr on the challenging MSRVTT video QA and video caption tasks with a far smaller model size and data scale. It also demonstrates strong zero-shot transferability on vision-language and video-language tasks. Code and models will be released in https://github.com/alibaba/AliceMind.

cs.CV↗

arXiv2023

FedRolex: Model-Heterogeneous Federated Learning with Rolling Sub-Model Extraction

Most cross-device federated learning (FL) studies focus on the model-homogeneous setting where the global server model and local client models are identical. However, such constraint not only excludes low-end clients who would otherwise make unique contributions to model training but also restrains clients from training large models due to on-device resource bottlenecks. In this work, we propose FedRolex, a partial training (PT)-based approach that enables model-heterogeneous FL and can train a global server model larger than the largest client model. At its core, FedRolex employs a rolling sub-model extraction scheme that allows different parts of the global server model to be evenly trained, which mitigates the client drift induced by the inconsistency between individual client models and server model architectures. We show that FedRolex outperforms state-of-the-art PT-based model-heterogeneous FL methods (e.g. Federated Dropout) and reduces the gap between model-heterogeneous and model-homogeneous FL, especially under the large-model large-dataset regime. In addition, we provide theoretical statistical analysis on its advantage over Federated Dropout and evaluate FedRolex on an emulated real-world device distribution to show that FedRolex can enhance the inclusiveness of FL and boost the performance of low-end devices that would otherwise not benefit from FL. Our code is available at: https://github.com/AIoT-MLSys-Lab/FedRolex

cs.LG↗

arXiv2023

A Comprehensive Study on Optimizing Systems with Data Processing Units

New hardware, such as SmartNICs, has been released to offload network applications in data centers. Off-path SmartNICs, a type of multi-core SoC SmartNICs, have attracted the attention of many researchers. Unfortunatelly, they lack the fully exploration of off-path SmartNICs. In this paper, we use a BlueField SmartNIC as an example to conduct a systematical study on the advantages and disadvantages of off-path SmartNICs. We make a detailed performance characterization on an off-path SmartNIC including computing power and network communication overhead, and propose the following advices: 1) Directly utilize the specific accelerators on the SmartNIC to offload applications; 2) Offload latency-insensitive background processing to the SmartNIC to reduce the load on the host; 3) Regard the SmartNIC as a new endpoint in the network to expand the computing power and storage resources of the server host; 4) Avoid directly employing the design method for systems based on on-path SmartNICs. We apply these advices to several use cases and show the performance improvements.

cs.NI↗

arXiv2022

HiTeA: Hierarchical Temporal-Aware Video-Language Pre-training

Video-language pre-training has advanced the performance of various downstream video-language tasks. However, most previous methods directly inherit or adapt typical image-language pre-training paradigms to video-language pre-training, thus not fully exploiting the unique characteristic of video, i.e., temporal. In this paper, we propose a Hierarchical Temporal-Aware video-language pre-training framework, HiTeA, with two novel pre-training tasks for modeling cross-modal alignment between moments and texts as well as the temporal relations of video-text pairs. Specifically, we propose a cross-modal moment exploration task to explore moments in videos, which results in detailed video moment representation. Besides, the inherent temporal relations are captured by aligning video-text pairs as a whole in different time resolutions with multi-modal temporal relation exploration task. Furthermore, we introduce the shuffling test to evaluate the temporal reliance of datasets and video-language pre-training models. We achieve state-of-the-art results on 15 well-established video-language understanding and generation tasks, especially on temporal-oriented datasets (e.g., SSv2-Template and SSv2-Label) with 8.6% and 11.1% improvement respectively. HiTeA also demonstrates strong generalization ability when directly transferred to downstream tasks in a zero-shot manner. Models and demo will be available on ModelScope.

cs.CV↗

arXiv2022

Eye-tracking based classification of Mandarin Chinese readers with and without dyslexia using neural sequence models

Eye movements are known to reflect cognitive processes in reading, and psychological reading research has shown that eye gaze patterns differ between readers with and without dyslexia. In recent years, researchers have attempted to classify readers with dyslexia based on their eye movements using Support Vector Machines (SVMs). However, these approaches (i) are based on highly aggregated features averaged over all words read by a participant, thus disregarding the sequential nature of the eye movements, and (ii) do not consider the linguistic stimulus and its interaction with the reader's eye movements. In the present work, we propose two simple sequence models that process eye movements on the entire stimulus without the need of aggregating features across the sentence. Additionally, we incorporate the linguistic stimulus into the model in two ways -- contextualized word embeddings and manually extracted linguistic features. The models are evaluated on a Mandarin Chinese dataset containing eye movements from children with and without dyslexia. Our results show that (i) even for a logographic script such as Chinese, sequence models are able to classify dyslexia on eye gaze sequences, reaching state-of-the-art performance, and (ii) incorporating the linguistic stimulus does not help to improve classification performance.

cs.CL↗

arXiv2022

High-temperature ferromagnetism and strong $π$-conjugation feature in two-dimensional manganese tetranitride

Two-dimensional (2D) magnetic materials have attracted tremendous research interest because of the promising application in the next-generation microelectronic devices. Here, by the first-principles calculations, we propose a two-dimensional ferromagnetic material with high Curie temperature, manganese tetranitride MnN$_4$ monolayer, which is a square-planar lattice made up of only one layer of atoms. The structure is demonstrated to be stable by the phonon spectra and the molecular dynamic simulations, and the stability is ascribed to the $π$-d conjugation between $π$ orbital of N=N bond and Mn $d$ orbital. More interestingly, the MnN$_4$ monolayer displays robust 2D ferromagnetism, which originates from the strong exchange couplings between Mn atoms due to the $π$-d conjugation. The high critical temperature of 247 K is determined by solving the Heisenberg model with the Monte Carlo method.

cond-mat.mtrl-sci↗

arXiv2022

A Scheme to fabricate magnetic graphene-like cobalt nitride CoN4monolayer proposed by first-principles calculations

We propose a scheme to fabricate the cobalt nitride CoN4 monolayer, a magnetic graphene-like two-dimensional material, in which all Co and N atoms are in a plane. Under the pressure above 40 GPa, the bulk CoN4 is stabilized in a triclinic phase. With the pressure decreasing, the triclinic phase of CoN4 is transformed into an orthorhombic phase, and the latter is a layered compound with large interlayer spacing. At ambient condition, the weak interlayer couplings are so small that single CoN4 layer can be exfoliated by the mechanical method.

cond-mat.mtrl-sci↗

arXiv2022

Zero-shot Image Captioning by Anchor-augmented Vision-Language Space Alignment

CLIP (Contrastive Language-Image Pre-Training) has shown remarkable zero-shot transfer capabilities in cross-modal correlation tasks such as visual classification and image retrieval. However, its performance in cross-modal generation tasks like zero-shot image captioning remains unsatisfied. In this work, we discuss that directly employing CLIP for zero-shot image captioning relies more on the textual modality in context and largely ignores the visual information, which we call \emph{contextual language prior}. To address this, we propose Cross-modal Language Models (CLMs) to facilitate unsupervised cross-modal learning. We further propose Anchor Augment to guide the generative model's attention to the fine-grained information in the representation of CLIP. Experiments on MS COCO and Flickr 30K validate the promising performance of proposed approach in both captioning quality and computational efficiency.

cs.CV↗

arXiv2022

X-CLIP: End-to-End Multi-grained Contrastive Learning for Video-Text Retrieval

Video-text retrieval has been a crucial and fundamental task in multi-modal research. The development of video-text retrieval has been considerably promoted by large-scale multi-modal contrastive pre-training, which primarily focuses on coarse-grained or fine-grained contrast. However, cross-grained contrast, which is the contrast between coarse-grained representations and fine-grained representations, has rarely been explored in prior research. Compared with fine-grained or coarse-grained contrasts, cross-grained contrast calculate the correlation between coarse-grained features and each fine-grained feature, and is able to filter out the unnecessary fine-grained features guided by the coarse-grained feature during similarity calculation, thus improving the accuracy of retrieval. To this end, this paper presents a novel multi-grained contrastive model, namely X-CLIP, for video-text retrieval. However, another challenge lies in the similarity aggregation problem, which aims to aggregate fine-grained and cross-grained similarity matrices to instance-level similarity. To address this challenge, we propose the Attention Over Similarity Matrix (AOSM) module to make the model focus on the contrast between essential frames and words, thus lowering the impact of unnecessary frames and words on retrieval results. With multi-grained contrast and the proposed AOSM module, X-CLIP achieves outstanding performance on five widely-used video-text retrieval datasets, including MSR-VTT (49.3 R@1), MSVD (50.4 R@1), LSMDC (26.1 R@1), DiDeMo (47.8 R@1) and ActivityNet (46.2 R@1). It outperforms the previous state-of-theart by +6.3%, +6.6%, +11.1%, +6.7%, +3.8% relative improvements on these benchmarks, demonstrating the superiority of multi-grained contrast and AOSM.

cs.CV↗

arXiv2022

Asymptotic behaviour of rotating convection-driven dynamos in the plane layer geometry

Dynamos driven by rotating convection in the plane layer geometry are investigated numerically for a range of Ekman number ($E$), magnetic Prandtl number ($Pm$) and Rayleigh number ($Ra$). The primary purpose of the investigation is to compare results of the simulations with previously developed asymptotic theory that is applicable in the limit of rapid rotation. We find that all of the simulations are in the quasi-geostrophic regime in which the Coriolis and pressure gradient forces are approximately balanced at leading order, whereas all other forces, including the Lorentz force, act as perturbations. Agreement between simulation output and asymptotic scalings for the energetics, flow speeds, magnetic field amplitude and length scales is found. The transition from large scale dynamos to small scale dynamos is well described by the magnetic Reynolds number based on the small convective length scale, $\widetilde{Rm}$, with large scale dynamos preferred when $\widetilde{Rm} \lesssim O(1)$. The magnitude of the large scale magnetic field is observed to saturate and become approximately constant with increasing Rayleigh number. Energy spectra show that all length scales present in the flow field and the small-scale magnetic field are consistent with a scaling of $E^{1/3}$, even in the turbulent regime. For a fixed value of $E$, we find that the viscous dissipation length scale is approximately constant over a broad range of $Ra$; the ohmic dissipation length scale is approximately constant within the large scale dynamo regime, but transitions to a $\widetilde{Rm}^{-1/2}$ scaling in the small scale dynamo regime.

physics.flu-dyn↗

arXiv2022

DictBERT: Dictionary Description Knowledge Enhanced Language Model Pre-training via Contrastive Learning

Although pre-trained language models (PLMs) have achieved state-of-the-art performance on various natural language processing (NLP) tasks, they are shown to be lacking in knowledge when dealing with knowledge driven tasks. Despite the many efforts made for injecting knowledge into PLMs, this problem remains open. To address the challenge, we propose \textbf{DictBERT}, a novel approach that enhances PLMs with dictionary knowledge which is easier to acquire than knowledge graph (KG). During pre-training, we present two novel pre-training tasks to inject dictionary knowledge into PLMs via contrastive learning: \textit{dictionary entry prediction} and \textit{entry description discrimination}. In fine-tuning, we use the pre-trained DictBERT as a plugin knowledge base (KB) to retrieve implicit knowledge for identified entries in an input sequence, and infuse the retrieved knowledge into the input to enhance its representation via a novel extra-hop attention mechanism. We evaluate our approach on a variety of knowledge driven and language understanding tasks, including NER, relation extraction, CommonsenseQA, OpenBookQA and GLUE. Experimental results demonstrate that our model can significantly improve typical PLMs: it gains a substantial improvement of 0.5\%, 2.9\%, 9.0\%, 7.1\% and 3.3\% on BERT-large respectively, and is also effective on RoBERTa-large.

cs.CL↗

arXiv2022

Self-Referenced Terahertz Semiconductor Dual-Comb Sources

Employing two frequency combs with a slight difference in repetition frequencies, the dual-comb source shows unique advantages in high precision spectroscopy, imaging, ranging, communications, etc. In the terahertz (THz) frequency range, the electrically pumped quantum cascade laser (QCL) offers the possibility of realizing the compact dual-comb source due to its semiconductor-based chip-scale configuration. Although the active stabilization of a THz QCL dual-comb source was demonstrated by phase locking one of the dual-comb lines, the full stabilization of all dual-comb lines is still challenging. Here, we propose a self-reference method to obtain a fully stabilized dual-comb signal on a pure THz QCL platform. Without using any external locking components, we filter out one dual-comb line and beat it with the whole dual-comb signal, which eliminates the common carrier offset frequency noise and reduces the dual-comb repetition frequency noise. It is experimentally demonstrated that the self-reference technique can significantly improve the long-term stability of the dual-comb signal. A record of the ``maxhold" linewidth of 14.8 kHz (60 s time duration) is obtained by implementing the self-reference technique, while without the self-reference the dual-comb lines show a ``maxhold" linewidth of 2 MHz (15 s time duration). The method provides the simplest way to improve the long-term stability of THz QCL dual-comb sources, which can be further adopted for high precision measurements.

physics.optics↗

arXiv2022

mPLUG: Effective and Efficient Vision-Language Learning by Cross-modal Skip-connections

Large-scale pretrained foundation models have been an emerging paradigm for building artificial intelligence (AI) systems, which can be quickly adapted to a wide range of downstream tasks. This paper presents mPLUG, a new vision-language foundation model for both cross-modal understanding and generation. Most existing pre-trained models suffer from the problems of low computational efficiency and information asymmetry brought by the long visual sequence in cross-modal alignment. To address these problems, mPLUG introduces an effective and efficient vision-language architecture with novel cross-modal skip-connections, which creates inter-layer shortcuts that skip a certain number of layers for time-consuming full self-attention on the vision side. mPLUG is pre-trained end-to-end on large-scale image-text pairs with both discriminative and generative objectives. It achieves state-of-the-art results on a wide range of vision-language downstream tasks, such as image captioning, image-text retrieval, visual grounding and visual question answering. mPLUG also demonstrates strong zero-shot transferability when directly transferred to multiple video-language tasks.

cs.CL↗

arXiv2022

WikiDiverse: A Multimodal Entity Linking Dataset with Diversified Contextual Topics and Entity Types

Multimodal Entity Linking (MEL) which aims at linking mentions with multimodal contexts to the referent entities from a knowledge base (e.g., Wikipedia), is an essential task for many multimodal applications. Although much attention has been paid to MEL, the shortcomings of existing MEL datasets including limited contextual topics and entity types, simplified mention ambiguity, and restricted availability, have caused great obstacles to the research and application of MEL. In this paper, we present WikiDiverse, a high-quality human-annotated MEL dataset with diversified contextual topics and entity types from Wikinews, which uses Wikipedia as the corresponding knowledge base. A well-tailored annotation procedure is adopted to ensure the quality of the dataset. Based on WikiDiverse, a sequence of well-designed MEL models with intra-modality and inter-modality attentions are implemented, which utilize the visual information of images more adequately than existing MEL models do. Extensive experimental analyses are conducted to investigate the contributions of different modalities in terms of MEL, facilitating the future research on this task. The dataset and baseline models are available at https://github.com/wangxw5/wikiDiverse.

cs.CL↗

arXiv2022

Shifting More Attention to Visual Backbone: Query-modulated Refinement Networks for End-to-End Visual Grounding

Visual grounding focuses on establishing fine-grained alignment between vision and natural language, which has essential applications in multimodal reasoning systems. Existing methods use pre-trained query-agnostic visual backbones to extract visual feature maps independently without considering the query information. We argue that the visual features extracted from the visual backbones and the features really needed for multimodal reasoning are inconsistent. One reason is that there are differences between pre-training tasks and visual grounding. Moreover, since the backbones are query-agnostic, it is difficult to completely avoid the inconsistency issue by training the visual backbone end-to-end in the visual grounding framework. In this paper, we propose a Query-modulated Refinement Network (QRNet) to address the inconsistent issue by adjusting intermediate features in the visual backbone with a novel Query-aware Dynamic Attention (QD-ATT) mechanism and query-aware multiscale fusion. The QD-ATT can dynamically compute query-dependent visual attention at the spatial and channel levels of the feature maps produced by the visual backbone. We apply the QRNet to an end-to-end visual grounding framework. Extensive experiments show that the proposed method outperforms state-of-the-art methods on five widely used datasets.

cs.CV↗