arXiv subjects

Yong Zhang

Publications and source records attributed to Yong Zhang.

At least 433 records · Page 24Linked to original sources

arXiv2022

Cross-subject Action Unit Detection with Meta Learning and Transformer-based Relation Modeling

Facial Action Unit (AU) detection is a crucial task for emotion analysis from facial movements. The apparent differences of different subjects sometimes mislead changes brought by AUs, resulting in inaccurate results. However, most of the existing AU detection methods based on deep learning didn't consider the identity information of different subjects. The paper proposes a meta-learning-based cross-subject AU detection model to eliminate the identity-caused differences. Besides, a transformer-based relation learning module is introduced to learn the latent relations of multiple AUs. To be specific, our proposed work is composed of two sub-tasks. The first sub-task is meta-learning-based AU local region representation learning, called MARL, which learns discriminative representation of local AU regions that incorporates the shared information of multiple subjects and eliminates identity-caused differences. The second sub-task uses the local region representation of AU of the first sub-task as input, then adds relationship learning based on the transformer encoder architecture to capture AU relationships. The entire training process is cascaded. Ablation study and visualization show that our MARL can eliminate identity-caused differences, thus obtaining a robust and generalized AU discriminative embedding representation. Our results prove that on the two public datasets BP4D and DISFA, our method is superior to the state-of-the-art technology, and the F1 score is improved by 1.3% and 1.4%, respectively.

cs.CV↗

arXiv2022

The effect of dressing on thermalization of interacting waves

We propose a more general setup for prethermalization in the system of interacting waves. The idea lies in dividing the multi-wave interactions into trivial and nontrivial ones. The trivial interactions will dress waves and lead to a less strongly interacting system which is statistically equivalent to the original one. With this in mind, we find that prethermalization occurs not only in the weakly interacting regime but also in the strongly interacting regime. The irreversible process towards equilibrium is governed by the Zakharov equation, from which double scaling of the thermalization time is expected. Finally, the theory is well confirmed in numerical experiments.

cond-mat.stat-mech↗

arXiv2022

FAST search for circumstellar atomic hydrogen--I: the young planetary nebula IC 4997

Using the Five-hundred-meter Aperture Spherical radio Telescope (FAST) in Guizhou, China, we detect the 21cm neutral atomic hydrogen absorption in the young planetary nebula IC 4997. The absorption arises from a shell also associated with Na I D lines. The H I shell has a mass of $1.46\times10^{-2}$ M$_\odot$ and a dynamic age of 990yr. The column density of H I is estimated to be $7.1\times10^{20}$ cm$^{-2}$, which can be well explained in terms of a photodissociation region around the ionized nebula, limited by self shielding of H$_2$. We find that the atomic-to-ionized hydrogen ratio is 0.6, suggesting that H I substantially contributes to overall nebular mass.

astro-ph.SR↗

arXiv2022

A characteristic-spectral-mixed scheme for six-dimensional Wigner-Coulomb dynamics

Numerical resolution for 6-D Wigner dynamics under the Coulomb potential faces with the combined challenges of high dimensionality, nonlocality, oscillation and singularity. In particular, the extremely huge memory storage of 6-D grids hinders the usage of all existing deterministic numerical scheme, which is well-known as the curse of dimensionality. To surmount these difficulties, we propose a massively parallel solver, termed the CHAracteristic-Spectral-Mixed (CHASM) scheme, by fully exploiting two distinct features of the Wigner equation: Locality of spatial advection and nonlocality of quantum interaction. Our scheme utilizes the local cubic B-spline basis to interpolate the local spatial advection. The key is to use a perfectly matched boundary condition to give a closure of spline coefficients, so that distributed pieces can recover the global one as accurately as possible owing to the rapid decay of wavelet basis in the dual space, and communication costs are significantly reduced. To resolve the nonlocal pseudodifferential operator with weakly singular symbol, CHASM further adopts the truncated kernel method to attain a highly efficient approximation. Several typical experiments including the quantum harmonic oscillator and Hydrogen 1s state demonstrate the accuracy and efficiency of CHASM. The non-equilibrium electron-proton couplings are also clearly displayed and reveal the uncertainty principle and quantum tunneling in phase space. Finally, the scalability of CHASM up to 16000 cores is presented.

quant-ph↗

arXiv2022

Performance evaluations on the parallel CHAracteristic-Spectral-Mixed (CHASM) scheme

Performance evaluations on the deterministic algorithms for 6-D problems are rarely found in literatures except some recent advances in the Vlasov and Boltzmann community [Dimarco et al. (2018), Kormann et al. (2019)], due to the extremely high complexity. Thus a detailed comparison among various techniques shall be useful to the researchers in the related fields. We try to make a thorough evaluation on a parallel CHAracteristic-Spectral-Mixed (CHASM) scheme to support its usage. CHASM utilizes the cubic B-spline expansion in the spatial space and spectral expansion in the momentum space, which many potentially overcome the computational burden in solving classical and quantum kinetic equations in 6-D phase space. Our purpose is three-pronged. First, we would like show that by imposing some effective Hermite boundary conditions, the local cubic spline can approximate to the global one as accurately as possible. Second, we will illustrate the necessity of adopting the truncated kernel method in calculating the pseudodifferential operator with a singular symbol, since the widely used pseudo-spectral method [Ringhofer (1990)] might fail to properly tackle the singularity. Finally, we make a comparison among non-splitting Lawson schemes and Strang operator splitting. Our numerical results demonstrate the advantage of the one-stage Lawson predictor-corrector scheme over multi-stage ones as well as the splitting scheme in both accuracy and stability.

math.NA↗

arXiv2022

The symmetry for two class of steady stratified periodic water waves

In this paper, we mainly consider two class of travelling stratified periodic water waves, one with negative (or without) surface tension and the other with constant Bernoulli's function and stagnation points. We first establish the symmetry result for stratified water waves with negative (or without) surface tension, but without stagnation by using the modified maximum principle. Furthermore, the symmetry property of stratified water waves with constant Bernoulli's function and stagnation points is also obtained provided the monotonic property is known.

math.AP↗

arXiv2022

The anti-Fermi-Pasta-Ulam-Tsingou problem in one-dimensional diatomic lattices

We study the thermalization dynamics of one-dimensional diatomic lattices (which represents the simplest system possessing multi-branch phonons), exemplified by the famous Fermi-Pasta-Ulam-Tsingou (FPUT)-$β$ and the Toda models. Here we focus on how the system relaxes to the equilibrium state when part of highest-frequency optical modes are initially excited, which is called the anti-FPUT problem comparing with the original FPUT problem (low frequency excitations of the monatomic lattice). It is shown numerically that the final thermalization time $T_{\rm eq}$ of the diatomic FPUT-$β$ chain depends on whether its acoustic modes are thermalized, whereas the $T_{\rm eq}$ of the diatomic Toda chain depends on the optical ones; in addition, the metastable state of both models have different energy distributions and lifetimes. Despite these differences, in the near-integrable region, the $T_{\rm eq}$ of both models still follows the same scaling law, i.e., $T_{\rm eq}$ is inversely proportional to the square of the perturbation strength. Finally, comparisons of the thermalization behavior between different models under various initial conditions are briefly summarized.

cond-mat.stat-mech↗

arXiv2022

VDTR: Video Deblurring with Transformer

Video deblurring is still an unsolved problem due to the challenging spatio-temporal modeling process. While existing convolutional neural network-based methods show a limited capacity for effective spatial and temporal modeling for video deblurring. This paper presents VDTR, an effective Transformer-based model that makes the first attempt to adapt Transformer for video deblurring. VDTR exploits the superior long-range and relation modeling capabilities of Transformer for both spatial and temporal modeling. However, it is challenging to design an appropriate Transformer-based model for video deblurring due to the complicated non-uniform blurs, misalignment across multiple frames and the high computational costs for high-resolution spatial modeling. To address these problems, VDTR advocates performing attention within non-overlapping windows and exploiting the hierarchical structure for long-range dependencies modeling. For frame-level spatial modeling, we propose an encoder-decoder Transformer that utilizes multi-scale features for deblurring. For multi-frame temporal modeling, we adapt Transformer to fuse multiple spatial features efficiently. Compared with CNN-based methods, the proposed method achieves highly competitive results on both synthetic and real-world video deblurring benchmarks, including DVD, GOPRO, REDS and BSD. We hope such a Transformer-based architecture can serve as a powerful alternative baseline for video deblurring and other video restoration tasks. The source code will be available at \url{https://github.com/ljzycmd/VDTR}.

cs.CV↗

arXiv2022

Molecules in the carbon-rich protoplanetary nebula CRL 2688

We present observations of the carbon-rich protoplanetary nebula (PPN) CRL 2688 made with the Institut de Radioastronomie Millimetrique (IRAM) 30 m telescope in the 3mm and 2mm bands. In total, 196 transition lines belonging to 38 molecular species and isotopologues are detected, among which, to our best knowledge, 153 transition lines and 13 species are the first report for this object. Additionally, in order to contribute to future research, we have collected observational data on the molecular lines of CRL 2688 from the literature and compiled them into a single unified catalog. We find that the molecular abundance of CRL 2688 cannot be explained by the standard model of a circumstellar envelope. The implications of metal-bearing molecules on circumstellar chemistry are discussed.

astro-ph.SR↗

arXiv2022

Self-supervised Learning of Adversarial Example: Towards Good Generalizations for Deepfake Detection

Recent studies in deepfake detection have yielded promising results when the training and testing face forgeries are from the same dataset. However, the problem remains challenging when one tries to generalize the detector to forgeries created by unseen methods in the training dataset. This work addresses the generalizable deepfake detection from a simple principle: a generalizable representation should be sensitive to diverse types of forgeries. Following this principle, we propose to enrich the "diversity" of forgeries by synthesizing augmented forgeries with a pool of forgery configurations and strengthen the "sensitivity" to the forgeries by enforcing the model to predict the forgery configurations. To effectively explore the large forgery augmentation space, we further propose to use the adversarial training strategy to dynamically synthesize the most challenging forgeries to the current model. Through extensive experiments, we show that the proposed strategies are surprisingly effective (see Figure 1), and they could achieve superior performance than the current state-of-the-art methods. Code is available at \url{https://github.com/liangchen527/SLADD}.

cs.CV↗

arXiv2022

Room Temperature Gate Tunable Non Reciprocal Charge Transport in Lattice Matched InSb/CdTe Heterostructures

The manipulation of symmetry provides an effective way to tailor the physical orders in solid-state systems. With the breaking of both the inversion and time-reversal symmetries, non-reciprocal magneto-transport may emerge in assorted non-magnetic systems to enrich spintronic physics. Here, we report the observation of the uni-directional magneto-resistance (UMR) in the lattice-matched InSb/CdTe film up to room temperature. Benefiting from the strong built-in electric field of $0.13 \mathrm{~V} \cdot \mathrm{nm}^{-1}$ in the hetero-junction region, the resulting Rashba-type spin-orbit coupling and quantum confinement warrant stable angular-dependent second-order charge current with the non-reciprocal coefficient 1-2 orders of magnitude larger than most non-centrosymmetric materials at 298 K. More importantly, this heterostructure configuration enables highly-efficient gate tuning of the rectification response in which the enhancement of the UMR amplitude by 40% is realized. Our results advocate the narrow-gap semiconductor-based hybrid system with the robust two-dimensional interfacial spin texture as a suitable platform for the pursuit of controllable chiral spin-orbit devices and applications.

cond-mat.mes-hall↗

arXiv2022

FENeRF: Face Editing in Neural Radiance Fields

Previous portrait image generation methods roughly fall into two categories: 2D GANs and 3D-aware GANs. 2D GANs can generate high fidelity portraits but with low view consistency. 3D-aware GAN methods can maintain view consistency but their generated images are not locally editable. To overcome these limitations, we propose FENeRF, a 3D-aware generator that can produce view-consistent and locally-editable portrait images. Our method uses two decoupled latent codes to generate corresponding facial semantics and texture in a spatial aligned 3D volume with shared geometry. Benefiting from such underlying 3D representation, FENeRF can jointly render the boundary-aligned image and semantic mask and use the semantic mask to edit the 3D volume via GAN inversion. We further show such 3D representation can be learned from widely available monocular image and semantic mask pairs. Moreover, we reveal that joint learning semantics and texture helps to generate finer geometry. Our experiments demonstrate that FENeRF outperforms state-of-the-art methods in various face editing tasks.

cs.CV↗

arXiv2022

StyleHEAT: One-Shot High-Resolution Editable Talking Face Generation via Pre-trained StyleGAN

One-shot talking face generation aims at synthesizing a high-quality talking face video from an arbitrary portrait image, driven by a video or an audio segment. One challenging quality factor is the resolution of the output video: higher resolution conveys more details. In this work, we investigate the latent feature space of a pre-trained StyleGAN and discover some excellent spatial transformation properties. Upon the observation, we explore the possibility of using a pre-trained StyleGAN to break through the resolution limit of training datasets. We propose a novel unified framework based on a pre-trained StyleGAN that enables a set of powerful functionalities, i.e., high-resolution video generation, disentangled control by driving video or audio, and flexible face editing. Our framework elevates the resolution of the synthesized talking face to 1024*1024 for the first time, even though the training dataset has a lower resolution. We design a video-based motion generation module and an audio-based one, which can be plugged into the framework either individually or jointly to drive the video generation. The predicted motion is used to transform the latent features of StyleGAN for visual animation. To compensate for the transformation distortion, we propose a calibration network as well as a domain loss to refine the features. Moreover, our framework allows two types of facial editing, i.e., global editing via GAN inversion and intuitive editing based on 3D morphable models. Comprehensive experiments show superior video quality, flexible controllability, and editability over state-of-the-art methods.

cs.CV↗

arXiv2022

LAS-AT: Adversarial Training with Learnable Attack Strategy

Adversarial training (AT) is always formulated as a minimax problem, of which the performance depends on the inner optimization that involves the generation of adversarial examples (AEs). Most previous methods adopt Projected Gradient Decent (PGD) with manually specifying attack parameters for AE generation. A combination of the attack parameters can be referred to as an attack strategy. Several works have revealed that using a fixed attack strategy to generate AEs during the whole training phase limits the model robustness and propose to exploit different attack strategies at different training stages to improve robustness. But those multi-stage hand-crafted attack strategies need much domain expertise, and the robustness improvement is limited. In this paper, we propose a novel framework for adversarial training by introducing the concept of "learnable attack strategy", dubbed LAS-AT, which learns to automatically produce attack strategies to improve the model robustness. Our framework is composed of a target network that uses AEs for training to improve robustness and a strategy network that produces attack strategies to control the AE generation. Experimental evaluations on three benchmark databases demonstrate the superiority of the proposed method. The code is released at https://github.com/jiaxiaojunQAQ/LAS-AT.

cs.CV↗

arXiv2022

Membership Privacy Protection for Image Translation Models via Adversarial Knowledge Distillation

Image-to-image translation models are shown to be vulnerable to the Membership Inference Attack (MIA), in which the adversary's goal is to identify whether a sample is used to train the model or not. With daily increasing applications based on image-to-image translation models, it is crucial to protect the privacy of these models against MIAs. We propose adversarial knowledge distillation (AKD) as a defense method against MIAs for image-to-image translation models. The proposed method protects the privacy of the training samples by improving the generalizability of the model. We conduct experiments on the image-to-image translation models and show that AKD achieves the state-of-the-art utility-privacy tradeoff by reducing the attack performance up to 38.9% compared with the regular training model at the cost of a slight drop in the quality of the generated output images. The experimental results also indicate that the models trained by AKD generalize better than the regular training models. Furthermore, compared with existing defense methods, the results show that at the same privacy protection level, image translation models trained by AKD generate outputs with higher quality; while at the same quality of outputs, AKD enhances the privacy protection over 30%.

cs.CV↗

arXiv2022

Effect of surface H$_2$ on molecular hydrogen formation on interstellar grains

We investigate how the existence of hydrogen molecules on grain surfaces may affect H$_2$ formation efficiency in diffuse and translucent clouds. Hydrogen molecules are able to reduce the desorption energy of H atoms on grain surfaces in models. The detailed microscopic Monte Carlo method is used to perform model simulations. We found that the impact of the existence of H$_2$ on H$_2$ formation efficiency strongly depends on the diffusion barriers of H$_2$ on grain surfaces. Diffuse cloud models that do not consider surface H$_2$ predict that H atom recombination efficiency is above 0.5 over a grain temperature (T) range 10 K and 14 K. The adopted H$_2$ diffusion barriers in diffuse cloud models that consider surface H$_2$ are 80$\%$ H$_2$ desorption energies so that H$_2$ can be trapped in stronger binding sites. Depending on model parameters, these diffuse cloud models predict that the recombination efficiency is between nearly 0 and 0.5 at 10 K $\leq$ T $\leq$ 14 K. Translucent cloud model results show that H$_2$ formation efficiency is not affected by the existence of surface H$_2$ if the adopted average H$_2$ diffusion barrier on grain surfaces is low (194 K) so that H$_2$ can diffuse rapidly on grain surfaces. However, the recombination efficiency can drop to below 0.002 at T $\geq$ 10 K if higher average H$_2$ diffusion barrier is used (255 K) in translucent cloud models.

astro-ph.GA↗

arXiv2022

Cosine Model Watermarking Against Ensemble Distillation

Many model watermarking methods have been developed to prevent valuable deployed commercial models from being stealthily stolen by model distillations. However, watermarks produced by most existing model watermarking methods can be easily evaded by ensemble distillation, because averaging the outputs of multiple ensembled models can significantly reduce or even erase the watermarks. In this paper, we focus on tackling the challenging task of defending against ensemble distillation. We propose a novel watermarking technique named CosWM to achieve outstanding model watermarking performance against ensemble distillation. CosWM is not only elegant in design, but also comes with desirable theoretical guarantees. Our extensive experiments on public data sets demonstrate the excellent performance of CosWM and its advantages over the state-of-the-art baselines.

cs.CR↗

arXiv2022

E-LANG: Energy-Based Joint Inferencing of Super and Swift Language Models

Building huge and highly capable language models has been a trend in the past years. Despite their great performance, they incur high computational cost. A common solution is to apply model compression or choose light-weight architectures, which often need a separate fixed-size model for each desirable computational budget, and may lose performance in case of heavy compression. This paper proposes an effective dynamic inference approach, called E-LANG, which distributes the inference between large accurate Super-models and light-weight Swift models. To this end, a decision making module routes the inputs to Super or Swift models based on the energy characteristics of the representations in the latent space. This method is easily adoptable and architecture agnostic. As such, it can be applied to black-box pre-trained models without a need for architectural manipulations, reassembling of modules, or re-training. Unlike existing methods that are only applicable to encoder-only backbones and classification tasks, our method also works for encoder-decoder structures and sequence-to-sequence tasks such as translation. The E-LANG performance is verified through a set of experiments with T5 and BERT backbones on GLUE, SuperGLUE, and WMT. In particular, we outperform T5-11B with an average computations speed-up of 3.3$\times$ on GLUE and 2.9$\times$ on SuperGLUE. We also achieve BERT-based SOTA on GLUE with 3.2$\times$ less computations. Code and demo are available in the supplementary materials.

cs.CL↗