arXiv subjects

Zhi Wang

Publications and source records attributed to Zhi Wang.

At least 271 records · Page 15Linked to original sources

arXiv2022

Fully Self-Supervised Learning for Semantic Segmentation

In this work, we present a fully self-supervised framework for semantic segmentation(FS^4). A fully bootstrapped strategy for semantic segmentation, which saves efforts for the huge amount of annotation, is crucial for building customized models from end-to-end for open-world domains. This application is eagerly needed in realistic scenarios. Even though recent self-supervised semantic segmentation methods have gained great progress, these works however heavily depend on the fully-supervised pretrained model and make it impossible a fully self-supervised pipeline. To solve this problem, we proposed a bootstrapped training scheme for semantic segmentation, which fully leveraged the global semantic knowledge for self-supervision with our proposed PGG strategy and CAE module. In particular, we perform pixel clustering and assignments for segmentation supervision. Preventing it from clustering a mess, we proposed 1) a pyramid-global-guided (PGG) training strategy to supervise the learning with pyramid image/patch-level pseudo labels, which are generated by grouping the unsupervised features. The stable global and pyramid semantic pseudo labels can prevent the segmentation from learning too many clutter regions or degrading to one background region; 2) in addition, we proposed context-aware embedding (CAE) module to generate global feature embedding in view of its neighbors close both in space and appearance in a non-trivial way. We evaluate our method on the large-scale COCO-Stuff dataset and achieved 7.19 mIoU improvements on both things and stuff objects

cs.CV↗

arXiv2022

Towards a Systematic Survey for Carbon Neutral Data Centers

Data centers are carbon-intensive enterprises due to their massive energy consumption, and it is estimated that data center industry will account for 8\% of global carbon emissions by 2030. However, both technological and policy instruments for reducing or even neutralizing data center carbon emissions have not been thoroughly investigated. To bridge this gap, this survey paper proposes a roadmap towards carbon-neutral data centers that takes into account both policy instruments and technological methodologies. We begin by presenting the carbon footprint of data centers, as well as some insights into the major sources of carbon emissions. Following that, carbon neutrality plans for major global cloud providers are discussed to summarize current industrial efforts in this direction. In what follows, we introduce the carbon market as a policy instrument to explain how to offset data center carbon emissions in a cost-efficient manner. On the technological front, we propose achieving carbon-neutral data centers by increasing renewable energy penetration, improving energy efficiency, and boosting energy circulation simultaneously. A comprehensive review of existing technologies on these three topics is elaborated subsequently. Based on this, a multi-pronged approach towards carbon neutrality is envisioned and a digital twin-powered industrial artificial intelligence (AI) framework is proposed to make this solution a reality. Furthermore, three key scientific challenges for putting such a framework in place are discussed. Finally, several applications for this framework are presented to demonstrate its enormous potential.

cs.CY↗

arXiv2022

Provably Efficient Multi-Task Reinforcement Learning with Model Transfer

We study multi-task reinforcement learning (RL) in tabular episodic Markov decision processes (MDPs). We formulate a heterogeneous multi-player RL problem, in which a group of players concurrently face similar but not necessarily identical MDPs, with a goal of improving their collective performance through inter-player information sharing. We design and analyze an algorithm based on the idea of model transfer, and provide gap-dependent and gap-independent upper and lower bounds that characterize the intrinsic complexity of the problem.

cs.LG↗

arXiv2021

Partial Reduction and Cosmology at Defect Brane

Partial reduction is a Randall-Sundrum reduction for only part of the AdS region between finite tension brane and zero tension brane. This is interesting in AdS/BCFT where the AdS bulk contains a defect brane. We employ partial reduction for a AdS bulk with a brane evolving as a $2d$ Friedmann-Robertson-Walker (FRW) cosmology and demonstrate the equivalence between defect extremal surface and island formula for a large subregion fine grained entropy in boundary CFT. We then move to higher dimensions and demonstrate the existence of $4d$ massless graviton on AdS$_4$ brane in partial reduction. We also propose a partial reduction for a $4d$ FRW cosmology at defect brane and obtain the Newton constant by computing boundary entropy.

hep-th↗

arXiv2021

Photoinduced charge density wave transition like a puppet on a string

Charge density wave (CDW) materials can undergo an ultrafast phase transition after an ultrashort laser pulse excitation, and the suggested underlying mechanisms have always been associated with two main features: excitonic interaction-induced CDW charge order and electron-phonon coupling-induced periodic lattice distortion (PLD). Here, beyond these two mechanisms, we reveal that photoexcitation induced CDW phase transition in the prototypic CDW example 1T-TiSe2 is similar to a puppet on a string: six Ti-Se bonds connected to each distorted Ti atom acting as six strings controlling the PLD and in turn the CDW orders. The photoexcitation induced modulation on charge population of the Ti-Se bonds generates a laser-fluence-dependent interatomic-repulsive force along each Ti-Se bond. The nonequal length of these six Ti-Se bonds gives rise to a net force exerted on the central distorted Ti atom to push it toward the suppressing of the PLD and thus the CDW orders. We further illustrate that the dynamics of each distorted Ti atom behaves as though it attached to a spring in a simple harmonic motion with a fluence-dependent oscillation frequency, uniting two previously reported scaling laws for phase transition time. These findings significantly advance the understanding of CDW instability and provide new insights into how photoexcitation induced modulation on charge population may lead to phase transitions by directly connecting interatomic forces with reaction pathways.

cond-mat.mtrl-sci↗

arXiv2021

Uniting the order and disorder dynamics in photoexcited VO2

Photoinduced phase transition (PIPT) is always treated as a coherent process, but ultrafast disordering in PIPT is observed in recent experiments. Utilizing the real-time time-dependent density functional theory (rt-TDDFT) method, here, we track the motion of individual vanadium (V) ions during PIPT in VO2 and uncover that their coherent or disordered dynamics can be manipulated by tuning the laser fluence. We find that the photoexcited holes generate a force on each V-V dimer to drive their collective coherent motion, in competing with the thermal-induced vibrations. If the laser fluence is so weak that the photoexcited hole density is too low to drive the phase transition alone, the PIPT is a disordered process due to the interference of thermal phonons. We also reveal that the photoexcited holes populated by the V-V dimerized bonding states will become saturated if the laser fluence is too strong, limiting the timescale of photoinduced phase transition.

cond-mat.mtrl-sci↗

arXiv2021

Neurodynamic TDOA localization with NLOS mitigation via maximum correntropy criterion

In this paper, we exploit the maximum correntropy criterion (MCC) to robustify the traditional time-difference-of-arrival (TDOA) location estimator in the presence of non-line-of-sight (NLOS) propagation conditions. For the sake of statistical efficiency, the correntropy-based robust loss is imposed on the underlying time-of-arrival composition via joint estimation of the source position and onset time, instead of the TDOA counterpart generated in the postprocessing of sensor-collected timestamps. We then employ a neurodynamic optimization approach to tackle the highly nonconvex MCC formulation. Furthermore, we examine the local stability of equilibrium for the corresponding projection-type neural network model. Simulation investigations in representative NLOS propagation scenarios demonstrate that our neurodynamic robust TDOA localization solution is capable of outperforming several existing schemes in terms of positioning accuracy.

eess.SP↗

arXiv2021

bert2BERT: Towards Reusable Pretrained Language Models

In recent years, researchers tend to pre-train ever-larger language models to explore the upper limit of deep models. However, large language model pre-training costs intensive computational resources and most of the models are trained from scratch without reusing the existing pre-trained models, which is wasteful. In this paper, we propose bert2BERT, which can effectively transfer the knowledge of an existing smaller pre-trained model (e.g., BERT_BASE) to a large model (e.g., BERT_LARGE) through parameter initialization and significantly improve the pre-training efficiency of the large model. Specifically, we extend the previous function-preserving on Transformer-based language model, and further improve it by proposing advanced knowledge for large model's initialization. In addition, a two-stage pre-training method is proposed to further accelerate the training process. We did extensive experiments on representative PLMs (e.g., BERT and GPT) and demonstrate that (1) our method can save a significant amount of training cost compared with baselines including learning from scratch, StackBERT and MSLT; (2) our method is generic and applicable to different types of pre-trained models. In particular, bert2BERT saves about 45% and 47% computational cost of pre-training BERT_BASE and GPT_BASE by reusing the models of almost their half sizes. The source code will be publicly available upon publication.

cs.CL↗

arXiv2021

ECMO: Peripheral Transplantation to Rehost Embedded Linux Kernels

Dynamic analysis based on the full-system emulator QEMU is widely used for various purposes. However, it is challenging to run firmware images of embedded devices in QEMU, especially the process to boot the Linux kernel (we call this process rehosting the Linux kernel.) That's because embedded devices usually use different system-on-chips (SoCs) from multiple vendors and only a limited number of SoCs are currently supported in QEMU. In this work, we propose a technique called peripheral transplantation. The main idea is to transplant the device drivers of designated peripherals into the Linux kernel. By doing so, it can replace the peripherals in the kernel that are currently unsupported in QEMU with supported ones, thus making the Linux kernel rehostable. After that, various applications can be built upon. We implemented this technique inside a prototype system called ECMO and applied it to 815 firmware images, which consist of 20 kernel versions, 37 device models, and 24 vendors. The result shows that ECMO can successfully transplant peripherals for all the 815 Linux kernels. Among them,710 kernels can be successfully rehosted, i.e., launching a user-space shell (87.1% success rate). The failed cases are mainly because the root file system format (ramfs) is not supported by the kernel. We further build three applications, i.e., kernel crash analysis, rootkit forensic analysis, and kernel fuzzing, based on the rehosted kernels to demonstrate the usage scenarios of ECMO

cs.AR↗

arXiv2021

Accurate Assessment via Process Data

Accurate assessment of students' ability is the key task of a test. Assessments based on final responses are the standard. As the infrastructure advances, substantially more information is observed. One of such instances is the process data that is collected by computer-based interactive items, which contain a student's detailed interactive processes. In this paper, we show both theoretically and empirically that appropriately including such information in the assessment will substantially improve relevant assessment precision. The precision is measured empirically by out-of-sample test reliability.

stat.AP↗

arXiv2021

A Polarizable Water Potential Derived from a Model Electron Density

A new empirical potential for efficient, large scale molecular dynamics simulation of water is presented. The HIPPO (Hydrogen-like Intermolecular Polarizable POtential) force field is based upon the model electron density of a hydrogen-like atom. This framework is used to derive and parameterize individual terms describing charge penetration damped permanent electrostatics, damped polarization, charge transfer, anisotropic Pauli repulsion, and damped dispersion interactions. Initial parameter values were fit to Symmetry Adapted Perturbation Theory (SAPT) energy components for ten water dimer configurations, as well as the radial and angular dependence of the canonical dimer. The SAPT-based parameters were then systematically refined to extend the treatment to water bulk phases. The final HIPPO water model provides a balanced representation of a wide variety of properties of gas phase clusters, liquid water and ice polymorphs, across a range of temperatures and pressures. This water potential yields a rationalization of water structure, dynamics and thermodynamics explicitly correlated with an ab initio energy decomposition, while providing a level of accuracy comparable or superior to previous polarizable atomic multipole force fields. The HIPPO water model serves as a cornerstone around which similarly detailed physics-based models can be developed for additional molecular species.

physics.chem-ph↗

arXiv2021

A Closer Look at Temporal Sentence Grounding in Videos: Dataset and Metric

Temporal Sentence Grounding in Videos (TSGV), i.e., grounding a natural language sentence which indicates complex human activities in a long and untrimmed video sequence, has received unprecedented attentions over the last few years. Although each newly proposed method plausibly can achieve better performance than previous ones, current TSGV models still tend to capture the moment annotation biases and fail to take full advantage of multi-modal inputs. Even more incredibly, several extremely simple baselines without training can also achieve state-of-the-art performance. In this paper, we take a closer look at the existing evaluation protocols for TSGV, and find that both the prevailing dataset splits and evaluation metrics are the devils to cause unreliable benchmarking. To this end, we propose to re-organize two widely-used TSGV benchmarks (ActivityNet Captions and Charades-STA). Specifically, we deliberately make the ground-truth moment distribution different in the training and test splits, i.e., out-of-distribution (OOD) testing. Meanwhile, we introduce a new evaluation metric dR@n,IoU@m to calibrate the basic IoU scores by penalizing on the bias-influenced moment predictions and alleviate the inflating evaluations caused by the dataset annotation biases such as overlong ground-truth moments. Under our new evaluation protocol, we conduct extensive experiments and ablation studies on eight state-of-the-art TSGV methods. All the results demonstrate that the re-organized dataset splits and new metric can better monitor the progress in TSGV. Our reorganized datsets are available at https://github.com/yytzsy/grounding_changing_distribution.

cs.CV↗

arXiv2021

A Survey on Temporal Sentence Grounding in Videos

Temporal sentence grounding in videos(TSGV), which aims to localize one target segment from an untrimmed video with respect to a given sentence query, has drawn increasing attentions in the research community over the past few years. Different from the task of temporal action localization, TSGV is more flexible since it can locate complicated activities via natural languages, without restrictions from predefined action categories. Meanwhile, TSGV is more challenging since it requires both textual and visual understanding for semantic alignment between two modalities(i.e., text and video). In this survey, we give a comprehensive overview for TSGV, which i) summarizes the taxonomy of existing methods, ii) provides a detailed description of the evaluation protocols(i.e., datasets and metrics) to be used in TSGV, and iii) in-depth discusses potential problems of current benchmarking designs and research directions for further investigations. To the best of our knowledge, this is the first systematic survey on temporal sentence grounding. More specifically, we first discuss existing TSGV approaches by grouping them into four categories, i.e., two-stage methods, end-to-end methods, reinforcement learning-based methods, and weakly supervised methods. Then we present the benchmark datasets and evaluation metrics to assess current research progress. Finally, we discuss some limitations in TSGV through pointing out potential problems improperly resolved in the current evaluation protocols, which may push forwards more cutting edge research in TSGV. Besides, we also share our insights on several promising directions, including three typical tasks with new and practical settings based on TSGV.

cs.CV↗

arXiv2021

Maximum correntropy criterion for robust TOA-based localization in NLOS environments

We investigate the problem of time-of-arrival (TOA) based localization under possible non-line-of-sight (NLOS) propagation conditions. To robustify the squared-range-based location estimator, we follow the maximum correntropy criterion, essentially the Welsch $M$-estimator with a redescending influence function which behaves like $\ell_0$-minimization towards the grossly biased measurements, to derive the formulation. The half-quadratic technique is then applied to settle the resulting optimization problem in an alternating maximization (AM) manner. By construction, the major computational challenge at each AM iteration boils down to handling an easily solvable generalized trust region subproblem. It is worth noting that the implementation of our localization method requires nothing but merely the TOA-based range measurements and sensor positions as prior information. Simulation and experimental results demonstrate the competence of the presented scheme in outperforming several state-of-the-art approaches in terms of positioning accuracy, especially in scenarios where the percentage of NLOS paths is not large enough.

eess.SP↗

arXiv2021

The emergent linear Rashba spin-orbit coupling offering the fast manipulation of hole-spin qubits in germanium

The electric dipole spin resonance (EDSR) combining strong spin-orbit coupling (SOC) and electric-dipole transitions facilitates fast spin control in a scalable way, which is the critical aspect of the rapid progress made recently in germanium (Ge) hole-spin qubits. However, a puzzle is raised because centrosymmetric Ge lacks the Dresselhaus SOC, a key element in the initial proposal of the hole-based EDSR. Here, we demonstrate that the recently uncovered finite k-linear Rashba SOC of 2D holes offers fast hole spin control via EDSR with Rabi frequencies in excellent agreement with experimental results over a wide range of driving fields. We also suggest that the Rabi frequency can reach 500 MHz under a higher gate electric field or multiple GHz in a replacement by [110]oriented wells. These findings bring a deeper understanding for hole-spin qubit manipulation and offer design principles to boost the gate speed.

cond-mat.mes-hall↗

arXiv2021

The effects of intrinsic local distortions vs. dynamic thermal motions on the stability and band gaps of cubic oxide and halide perovskites

Corner-shared ABX$_3$ perovskites have long featured prominently in solid-state chemistry and condensed matter physics. Still, the joint understanding of their two main subgroups-halides and oxides-has not been fully developed. Indeed, unlike the case that compounds having a single repeated motif ("monomorphous"), certain cubic perovskites can manifest a non-thermal distribution of local motifs ("polymorphous networks"). Such intrinsic deformations can include positional degrees of freedom. Unlike thermal motion, such intrinsic distortions do not time-average to zero. The present study compares electronic structure features of oxide and halide perovskites starting from the intrinsic polymorphous network described by DFT minimization of the internal energy, continuing to finite temperature thermal disorder using AIMD. We find that (i) different oxide vs. halide ABX$_3$ compounds adopt different energy-lowering distortion modes. The DFT calculated pair distribution function (PDF) of SrTiO$_3$ agrees with the recently measured PDF. (ii) In both oxides and halides, such intrinsic distortions lead to bandgap blueshifts with respect to monomorphous structure. (iii) For oxide perovskites, high-temperature AIMD simulations initiated from the polymorphous structures reveal that the thermally-induced distortions can lead to a bandgap redshift. (iv) In contrast, for cubic CsPbI$_3$, both the intrinsic distortions and the thermal distortions contribute in tandem to bandgap blueshift, the former, intrinsic effect being dominant. (v) In the oxide SrTiO$_3$ and CaTiO$_3$ (but not in halide), octahedral tilting leads to the emergence of a distinct $Γ$-$Γ$ direct bandgap component as a secondary valley minimum to the well-known indirect R-$Γ$ gap. Understanding such intrinsic vs. thermal effects on oxide vs. halide perovskites holds the potential for designing target electronic properties.

cond-mat.mtrl-sci↗

arXiv2021

Spin-orbit driven ferromagnetism at half moiré filling in magic-angle twisted bilayer graphene

Strong electron correlation and spin-orbit coupling (SOC) provide two non-trivial threads to condensed matter physics. When these two strands of physics come together, a plethora of quantum phenomena with novel topological order have been predicted to emerge in the correlated SOC regime. In this work, we examine the combined influence of electron correlation and SOC on a 2-dimensional (2D) electronic system at the atomic interface between magic-angle twisted bilayer graphene (tBLG) and a tungsten diselenide (\WSe) crystal. In such a structure, strong electron correlation within the moiré flatband stabilizes correlated insulating states at both quarter and half-filling, whereas SOC transforms these Mott-like insulators into ferromagnets, evidenced by robust anomalous Hall effect with hysteretic switching behavior. The coupling between spin and valley degrees of freedom is unambiguously demonstrated as the magnetic order is shown to be tunable with an in-plane magnetic field, or a perpendicular electric field. In addition, we examine the influence of SOC on the isospin order and stability of superconductivity. Our findings establish an efficient experimental knob to engineer topological properties of moiré bands in twisted bilayer graphene and related systems.

cond-mat.mes-hall↗

arXiv2021

EvilModel: Hiding Malware Inside of Neural Network Models

Delivering malware covertly and evasively is critical to advanced malware campaigns. In this paper, we present a new method to covertly and evasively deliver malware through a neural network model. Neural network models are poorly explainable and have a good generalization ability. By embedding malware in neurons, the malware can be delivered covertly, with minor or no impact on the performance of neural network. Meanwhile, because the structure of the neural network model remains unchanged, it can pass the security scan of antivirus engines. Experiments show that 36.9MB of malware can be embedded in a 178MB-AlexNet model within 1% accuracy loss, and no suspicion is raised by anti-virus engines in VirusTotal, which verifies the feasibility of this method. With the widespread application of artificial intelligence, utilizing neural networks for attacks becomes a forwarding trend. We hope this work can provide a reference scenario for the defense on neural network-assisted attacks.

cs.CR↗