arXiv subjects

Yi Zeng

Publications and source records attributed to Yi Zeng.

At least 199 records · Page 11Linked to original sources

arXiv2021

BackEISNN: A Deep Spiking Neural Network with Adaptive Self-Feedback and Balanced Excitatory-Inhibitory Neurons

Spiking neural networks (SNNs) transmit information through discrete spikes, which performs well in processing spatial-temporal information. Due to the non-differentiable characteristic, there still exist difficulties in designing well-performed SNNs. Recently, SNNs trained with backpropagation have shown superior performance due to the proposal of the gradient approximation. However, the performance on complex tasks is still far away from the deep neural networks. Taking inspiration from the autapse in the brain which connects the spiking neurons with a self-feedback connection, we apply an adaptive time-delayed self-feedback on the membrane potential to regulate the spike precisions. As well as, we apply the balanced excitatory and inhibitory neurons mechanism to control the spiking neurons' output dynamically. With the combination of the two mechanisms, we propose a deep spiking neural network with adaptive self-feedback and balanced excitatory and inhibitory neurons (BackEISNN). The experimental results on several standard datasets have shown that the two modules not only accelerate the convergence of the network but also improve the accuracy. For the MNIST, FashionMNIST, and N-MNIST datasets, our model has achieved state-of-the-art performance. For the CIFAR10 dataset, our BackEISNN also gets remarkable performance on a relatively light structure that competes against state-of-the-art SNNs.

cs.NE↗

arXiv2021

Fine-tuning Is Not Enough: A Simple yet Effective Watermark Removal Attack for DNN Models

Watermarking has become the tendency in protecting the intellectual property of DNN models. Recent works, from the adversary's perspective, attempted to subvert watermarking mechanisms by designing watermark removal attacks. However, these attacks mainly adopted sophisticated fine-tuning techniques, which have certain fatal drawbacks or unrealistic assumptions. In this paper, we propose a novel watermark removal attack from a different perspective. Instead of just fine-tuning the watermarked models, we design a simple yet powerful transformation algorithm by combining imperceptible pattern embedding and spatial-level transformations, which can effectively and blindly destroy the memorization of watermarked models to the watermark samples. We also introduce a lightweight fine-tuning strategy to preserve the model performance. Our solution requires much less resource or knowledge about the watermarking scheme than prior works. Extensive experimental results indicate that our attack can bypass state-of-the-art watermarking solutions with very high success rates. Based on our attack, we propose watermark augmentation techniques to enhance the robustness of existing watermarks.

cs.CR↗

arXiv2021

Fine and hyperfine interactions in $^{171}$YbOH and $^{173}$YbOH

The odd isotopologues of ytterbium monohydroxide, $^{171,173}$YbOH, have been identified as promising molecules in which to measure parity (P) and time reversal (T) violating physics. Here we characterize the $\tilde{A}^{2}Π_{1/2}(0,0,0)-\tilde{X}^2Σ^+(0,0,0)$ band near 577 nm for these odd isotopologues. Both laser-induced fluorescence (LIF) excitation spectra of a supersonic molecular beam sample and absorption spectra of a cryogenic buffer-gas cooled sample were recorded. Additionally, a novel spectroscopic technique based on laser-enhanced chemical reactions is demonstrated and utilized in the absorption measurements. This technique is especially powerful for disentangling congested spectra. An effective Hamiltonian model is used to extract the fine and hyperfine parameters for the $\tilde{A}^{2}Π_{1/2}(0,0,0)$ and $\tilde{X}^2Σ^+(0,0,0)$ states. A comparison of the determined $\tilde{X}^2Σ^+(0,0,0)$ hyperfine parameters with recently predicted values (M. Denis, et al., J. Chem. Phys. $\bf{152}$, 084303 (2020), K. Gaul and R. Berger, Phys. Rev. A $\bf{101}$, 012508 (2020), J. Liu et al., J. Chem. Phys. $\bf{154}$, 064110 (2021)) is made. The measured hyperfine parameters provide experimental confirmation of the computational methods used to compute the P,T-violating coupling constants $W_d$ and $W_M$, which correlate P,T-violating physics to P,T-violating energy shifts in the molecule. The dependence of the fine and hyperfine parameters of the $\tilde{A}^{2}Π_{1/2}(0,0,0)$ and $\tilde{X}^2Σ^+(0,0,0)$ states for all isotopologues of YbOH are discussed and a comparison to isoelectronic YbF is made.

physics.atom-ph↗

arXiv2021

DeepSweep: An Evaluation Framework for Mitigating DNN Backdoor Attacks using Data Augmentation

Public resources and services (e.g., datasets, training platforms, pre-trained models) have been widely adopted to ease the development of Deep Learning-based applications. However, if the third-party providers are untrusted, they can inject poisoned samples into the datasets or embed backdoors in those models. Such an integrity breach can cause severe consequences, especially in safety- and security-critical applications. Various backdoor attack techniques have been proposed for higher effectiveness and stealthiness. Unfortunately, existing defense solutions are not practical to thwart those attacks in a comprehensive way. In this paper, we investigate the effectiveness of data augmentation techniques in mitigating backdoor attacks and enhancing DL models' robustness. An evaluation framework is introduced to achieve this goal. Specifically, we consider a unified defense solution, which (1) adopts a data augmentation policy to fine-tune the infected model and eliminate the effects of the embedded backdoor; (2) uses another augmentation policy to preprocess input samples and invalidate the triggers during inference. We propose a systematic approach to discover the optimal policies for defending against different backdoor attacks by comprehensively evaluating 71 state-of-the-art data augmentation functions. Extensive experiments show that our identified policy can effectively mitigate eight different kinds of backdoor attacks and outperform five existing defense methods. We envision this framework can be a good benchmark tool to advance future DNN backdoor studies.

cs.CR↗

arXiv2021

SelfGait: A Spatiotemporal Representation Learning Method for Self-supervised Gait Recognition

Gait recognition plays a vital role in human identification since gait is a unique biometric feature that can be perceived at a distance. Although existing gait recognition methods can learn gait features from gait sequences in different ways, the performance of gait recognition suffers from insufficient labeled data, especially in some practical scenarios associated with short gait sequences or various clothing styles. It is unpractical to label the numerous gait data. In this work, we propose a self-supervised gait recognition method, termed SelfGait, which takes advantage of the massive, diverse, unlabeled gait data as a pre-training process to improve the representation abilities of spatiotemporal backbones. Specifically, we employ the horizontal pyramid mapping (HPM) and micro-motion template builder (MTB) as our spatiotemporal backbones to capture the multi-scale spatiotemporal representations. Experiments on CASIA-B and OU-MVLP benchmark gait datasets demonstrate the effectiveness of the proposed SelfGait compared with four state-of-the-art gait recognition methods. The source code has been released at https://github.com/EchoItLiu/SelfGait.

cs.CV↗

arXiv2020

Broadband inverted T-shaped seismic metamaterial

Seismic metamaterials (SMs) are expected to assist or replace traditional isolation systems owing to their strong attenuation of seismic waves. In this paper, a one-dimensional inverted T-shaped SM (1D ITSM) with an ultra-wide first bandgap (FBG) is proposed. The complex band structures are calculated to analyze the wave characteristics of the surface waves in the SMs. We find that the FBG of the 1D ITSM is composed of two parts; part 1 with surface evanescent waves and part 2 with no surface modes. Similar results are found in the complex band structure of the FBG of the SM consisting of periodically arranged pillars and substrate. The propagation of seismic surface waves in the 1D ITSM is different in these two frequency ranges of the FBG. In part 1, the seismic surface waves are significantly attenuated in the 1D ITSM, while in part 2, the surface waves are converted into bulk waves. Finally, the ultra-wide FBG is verified by using a kind of the two-dimensional ITSM in large-scale field experiments.

physics.app-ph↗

arXiv2020

FenceBox: A Platform for Defeating Adversarial Examples with Data Augmentation Techniques

It is extensively studied that Deep Neural Networks (DNNs) are vulnerable to Adversarial Examples (AEs). With more and more advanced adversarial attack methods have been developed, a quantity of corresponding defense solutions were designed to enhance the robustness of DNN models. It has become a popularity to leverage data augmentation techniques to preprocess input samples before inference to remove adversarial perturbations. By obfuscating the gradients of DNN models, these approaches can defeat a considerable number of conventional attacks. Unfortunately, advanced gradient-based attack techniques (e.g., BPDA and EOT) were introduced to invalidate these preprocessing effects. In this paper, we present FenceBox, a comprehensive framework to defeat various kinds of adversarial attacks. FenceBox is equipped with 15 data augmentation methods from three different categories. We comprehensively evaluated that these methods can effectively mitigate various adversarial attacks. FenceBox also provides APIs for users to easily deploy the defense over their models in different modes: they can either select an arbitrary preprocessing method, or a combination of functions for a better robustness guarantee, even under advanced adversarial attacks. We open-source FenceBox, and expect it can be used as a standard toolkit to facilitate the research of adversarial attacks and defenses.

cs.LG↗

arXiv2020

A Data Augmentation-based Defense Method Against Adversarial Attacks in Neural Networks

Deep Neural Networks (DNNs) in Computer Vision (CV) are well-known to be vulnerable to Adversarial Examples (AEs), namely imperceptible perturbations added maliciously to cause wrong classification results. Such variability has been a potential risk for systems in real-life equipped DNNs as core components. Numerous efforts have been put into research on how to protect DNN models from being tackled by AEs. However, no previous work can efficiently reduce the effects caused by novel adversarial attacks and be compatible with real-life constraints at the same time. In this paper, we focus on developing a lightweight defense method that can efficiently invalidate full whitebox adversarial attacks with the compatibility of real-life constraints. From basic affine transformations, we integrate three transformations with randomized coefficients that fine-tuned respecting the amount of change to the defended sample. Comparing to 4 state-of-art defense methods published in top-tier AI conferences in the past two years, our method demonstrates outstanding robustness and efficiency. It is worth highlighting that, our model can withstand advanced adaptive attack, namely BPDA with 50 rounds, and still helps the target model maintain an accuracy around 80 %, meanwhile constraining the attack success rate to almost zero.

cs.CR↗

arXiv2020

Mitigating Advanced Adversarial Attacks with More Advanced Gradient Obfuscation Techniques

Deep Neural Networks (DNNs) are well-known to be vulnerable to Adversarial Examples (AEs). A large amount of efforts have been spent to launch and heat the arms race between the attackers and defenders. Recently, advanced gradient-based attack techniques were proposed (e.g., BPDA and EOT), which have defeated a considerable number of existing defense methods. Up to today, there are still no satisfactory solutions that can effectively and efficiently defend against those attacks. In this paper, we make a steady step towards mitigating those advanced gradient-based attacks with two major contributions. First, we perform an in-depth analysis about the root causes of those attacks, and propose four properties that can break the fundamental assumptions of those attacks. Second, we identify a set of operations that can meet those properties. By integrating these operations, we design two preprocessing functions that can invalidate these powerful attacks. Extensive evaluations indicate that our solutions can effectively mitigate all existing standard and advanced attack techniques, and beat 11 state-of-the-art defense solutions published in top-tier conferences over the past 2 years. The defender can employ our solutions to constrain the attack success rate below 7% for the strongest attacks even the adversary has spent dozens of GPU hours.

cs.CR↗

arXiv2019

Responsible Facial Recognition and Beyond

Facial recognition is changing the way we live in and interact with our society. Here we discuss the two sides of facial recognition, summarizing potential risks and current concerns. We introduce current policies and regulations in different countries. Very importantly, we point out that the risks and concerns are not only from facial recognition, but also realistically very similar to other biometric recognition technology, including but not limited to gait recognition, iris recognition, fingerprint recognition, voice recognition, etc. To create a responsible future, we discuss possible technological moves and efforts that should be made to keep facial recognition (and biometric recognition in general) developing for social good.

cs.CV↗

arXiv2019

TEST: an End-to-End Network Traffic Examination and Identification Framework Based on Spatio-Temporal Features Extraction

With more encrypted network traffic gets involved in the Internet, how to effectively identify network traffic has become a top priority in the field. Accurate identification of the network traffic is the footstone of basic network services, say QoE, bandwidth allocation, and IDS. Previous identification methods either cannot deal with encrypted traffics or require experts to select tons of features to attain a relatively decent accuracy.In this paper, we present a Deep Learning based end-to-end network traffic identification framework, termed TEST, to avoid the aforementioned problems. CNN and LSTM are combined and implemented to help the machine automatically extract features from both special and time-related features of the raw traffic. The presented framework has two layers of structure, which made it possible to attain a remarkable accuracy on both encrypted traffic classification and intrusion detection tasks. The experimental results demonstrate that our model can outperform previous methods with a state-of-the-art accuracy of 99.98%.

cs.CR↗

arXiv2019

Towards High-Resolution Salient Object Detection

Deep neural network based methods have made a significant breakthrough in salient object detection. However, they are typically limited to input images with low resolutions ($400\times400$ pixels or less). Little effort has been made to train deep neural networks to directly handle salient object detection in very high-resolution images. This paper pushes forward high-resolution saliency detection, and contributes a new dataset, named High-Resolution Salient Object Detection (HRSOD). To our best knowledge, HRSOD is the first high-resolution saliency detection dataset to date. As another contribution, we also propose a novel approach, which incorporates both global semantic information and local high-resolution details, to address this challenging task. More specifically, our approach consists of a Global Semantic Network (GSN), a Local Refinement Network (LRN) and a Global-Local Fusion Network (GLFN). GSN extracts the global semantic information based on down-sampled entire image. Guided by the results of GSN, LRN focuses on some local regions and progressively produces high-resolution predictions. GLFN is further proposed to enforce spatial consistency and boost performance. Experiments illustrate that our method outperforms existing state-of-the-art methods on high-resolution saliency datasets by a large margin, and achieves comparable or even better performance than them on widely-used saliency benchmarks. The HRSOD dataset is available at https://github.com/yi94code/HRSOD.

cs.CV↗

arXiv2019

FlexNER: A Flexible LSTM-CNN Stack Framework for Named Entity Recognition

Named entity recognition (NER) is a foundational technology for information extraction. This paper presents a flexible NER framework compatible with different languages and domains. Inspired by the idea of distant supervision (DS), this paper enhances the representation by increasing the entity-context diversity without relying on external resources. We choose different layer stacks and sub-network combinations to construct the bilateral networks. This strategy can generally improve model performance on different datasets. We conduct experiments on five languages, such as English, German, Spanish, Dutch and Chinese, and biomedical fields, such as identifying the chemicals and gene/protein terms from scientific works. Experimental results demonstrate the good performance of this framework.

cs.CL↗

arXiv2019

A novel zero-frequency seismic metamaterial

A zero-frequency seismic metamaterial (ZFSM) consisting of a three-component seismic metamaterial plate and a half space is proposed to attenuate ultra-low frequency seismic surface waves. The design concept and models are verified firstly by lab-scale experiments on the seismic metamaterial consisting of a two-component seismic metamaterial plate and a half space. Then we calculate the band structures of the one-dimensional and two-dimensional ZFSMs, and evaluate their attenuation ability to Rayleigh waves. A wide band gap and a zero-frequency band gap (ZFBG) can be found as the band structure of the seismic metamaterial is almost equal to the band structure of the seismic metamaterial plate plus the sound cone. It is found that the Rayleigh waves in the ZFSM are deflected and converted into bulk waves. When the number of the unit cells of the ZFSM is sufficient, the transmission distance and deflection angle of the Rayleigh waves in the ZFSM are constant at the same frequency. This discovery is expected to open up the possibility of seismic protection for large nuclear power plants, ancient buildings and metropolitan areas.

physics.app-ph↗

arXiv2018

Linking Artificial Intelligence Principles

Artificial Intelligence principles define social and ethical considerations to develop future AI. They come from research institutes, government organizations and industries. All versions of AI principles are with different considerations covering different perspectives and making different emphasis. None of them can be considered as complete and can cover the rest AI principle proposals. Here we introduce LAIP, an effort and platform for linking and analyzing different Artificial Intelligence Principles. We want to explicitly establish the common topics and links among AI Principles proposed by different organizations and investigate on their uniqueness. Based on these efforts, for the long-term future of AI, instead of directly adopting any of the AI principles, we argue for the necessity of incorporating various AI Principles into a comprehensive framework and focusing on how they can interact and complete each other.

cs.AI↗

arXiv2018

The Fokker-Planck Equation for Lattice Vibration: Stochastic Dynamics and Thermal Conductivity

We propose a Fokker-Planck equation (FPE) theory to describe stochastic fluctuation and relaxation processes of lattice vibration at a wide range of conditions, including those beyond the phonon gas (PG) limit. Using the time-dependent, multiple state-variable probability function of a FPE, we first derive time-correlation functions of lattice heat currents in terms of correlation functions among multiple vibrational modes, and subsequently predict the thermal conductivity based on the Green-Kubo formalism. When the kinetic transport theories are valid, this FPE not only predicts a thermal conductivity that is identical to the one predicted by the phonon Boltzmann transport equation, but also provides additional microscopic details on the multiple-mode correlation functions. More importantly, when the kinetic theories become insufficient due to the breakdown of the PG approximation, this FPE theory remains valid to study the correlation functions among vibrational modes in highly anharmonic lattices with significant mode-mode interactions and/or in disordered lattices with strongly localized modes. At the limit of weak mode-mode interactions, we can adopt quantum perturbation theories to derive the drift/diffusion coefficients based on the anharmonicity data derived from first-principles methods. As temperature elevates to the classical regime, we can perform molecular dynamics simulations to directly compute the coefficients. Because these coefficients are defined as ensemble averages at the short time limit, we can implement massive parallel simulation algorithms to take full advantage of the paralleled high-performance computing platforms. A better understanding of the temperature-dependent coefficients up to melting temperatures will provide new insights on microscopic mechanisms that govern the heat conduction through anharmonic and/or disordered lattices beyond the PG model.

cond-mat.stat-mech↗

arXiv2014

Simple Formula for Marcus-Hush-Chidsey Kinetics

The Marcus-Hush-Chidsey (MHC) model is well known in electro-analytical chemistry as a successful microscopic theory of outer-sphere electron transfer at metal electrodes, but it is unfamiliar and rarely used in electrochemical engineering. One reason may be the difficulty of evaluating the MHC reaction rate, which is defined as an improper integral of the Marcus rate over the Fermi distribution of electron energies. Here, we report a simple analytical approximation of the MHC integral that interpolates between exact asymptotic limits for large overpotentials, as well as for large or small reorganization energies, and exhibits less than 5\% relative error for all reasonable parameter values. This result enables the MHC model to be considered as a practical alternative to the ubiquitous Butler-Volmer equation for improved understanding and engineering of electrochemical systems.

physics.chem-ph↗

arXiv2014

Phase Separation Dynamics in Isotropic Ion-Intercalation Particles

Lithium-ion batteries exhibit complex nonlinear dynamics, resulting from diffusion and phase transformations coupled to ion intercalation reactions. Using the recently developed Cahn-Hilliard reaction (CHR) theory, we investigate a simple mathematical model of ion intercalation in a spherical solid nanoparticle, which predicts transitions from solid-solution radial diffusion to two-phase shrinking-core dynamics. This general approach extends previous Li-ion battery models, which either neglect phase separation or postulate a spherical shrinking-core phase boundary, by predicting phase separation only under appropriate circumstances. The effect of the applied current is captured by generalized Butler-Volmer kinetics, formulated in terms of diffusional chemical potentials, and the model consistently links the evolving concentration profile to the battery voltage. We examine sources of charge/discharge asymmetry, such as asymmetric charge transfer and surface "wetting" by ions within the solid, which can lead to three distinct phase regions. In order to solve the fourth-order nonlinear CHR initial-boundary-value problem, a control-volume discretization is developed in spherical coordinates. The basic physics are illustrated by simulating many representative cases, including a simple model of the popular cathode material, lithium iron phosphate (neglecting crystal anisotropy and coherency strain). Analytical approximations are also derived for the voltage plateau as a function of the applied current.

physics.chem-ph↗