arXiv subjects

Cheng He

Publications and source records attributed to Cheng He.

At least 37 records · Page 2Linked to original sources

arXiv2022

SoloGAN: Multi-domain Multimodal Unpaired Image-to-Image Translation via a Single Generative Adversarial Network

Despite significant advances in image-to-image (I2I) translation with generative adversarial networks (GANs), it remains challenging to effectively translate an image to a set of diverse images in multiple target domains using a single pair of generator and discriminator. Existing I2I translation methods adopt multiple domain-specific content encoders for different domains, where each domain-specific content encoder is trained with images from the same domain only. Nevertheless, we argue that the content (domain-invariance) features should be learned from images among all of the domains. Consequently, each domain-specific content encoder of existing schemes fails to extract the domain-invariant features efficiently. To address this issue, we present a flexible and general SoloGAN model for efficient multimodal I2I translation among multiple domains with unpaired data. In contrast to existing methods, the SoloGAN algorithm uses a single projection discriminator with an additional auxiliary classifier and shares the encoder and generator for all domains. Consequently, the SoloGAN can be trained effectively with images from all domains such that the domain-invariance content representation can be efficiently extracted. Qualitative and quantitative results over a wide range of datasets against several counterparts and variants of the SoloGAN demonstrate the merits of the method, especially for challenging I2I translation datasets, i.e., datasets involving extreme shape variations or need to keep the complex backgrounds unchanged after translations. Furthermore, we demonstrate the contribution of each component in SoloGAN by ablation studies.

cs.CV↗

arXiv2021

Accelerating Multi-Objective Neural Architecture Search by Random-Weight Evaluation

For the goal of automated design of high-performance deep convolutional neural networks (CNNs), Neural Architecture Search (NAS) methodology is becoming increasingly important for both academia and industries.Due to the costly stochastic gradient descent (SGD) training of CNNs for performance evaluation, most existing NAS methods are computationally expensive for real-world deployments. To address this issue, we first introduce a new performance estimation metric, named Random-Weight Evaluation (RWE) to quantify the quality of CNNs in a cost-efficient manner. Instead of fully training the entire CNN, the RWE only trains its last layer and leaves the remainders with randomly initialized weights, which results in a single network evaluation in seconds.Second, a complexity metric is adopted for multi-objective NAS to balance the model size and performance. Overall, our proposed method obtains a set of efficient models with state-of-the-art performance in two real-world search spaces. Then the results obtained on the CIFAR-10 dataset are transferred to the ImageNet dataset to validate the practicality of the proposed algorithm. Moreover, ablation studies on NAS-Bench-301 datasets reveal the effectiveness of the proposed RWE in estimating the performance compared with existing methods.

cs.LG↗

arXiv2021

FaPN: Feature-aligned Pyramid Network for Dense Image Prediction

Recent advancements in deep neural networks have made remarkable leap-forwards in dense image prediction. However, the issue of feature alignment remains as neglected by most existing approaches for simplicity. Direct pixel addition between upsampled and local features leads to feature maps with misaligned contexts that, in turn, translate to mis-classifications in prediction, especially on object boundaries. In this paper, we propose a feature alignment module that learns transformation offsets of pixels to contextually align upsampled higher-level features; and another feature selection module to emphasize the lower-level features with rich spatial details. We then integrate these two modules in a top-down pyramidal architecture and present the Feature-aligned Pyramid Network (FaPN). Extensive experimental evaluations on four dense prediction tasks and four datasets have demonstrated the efficacy of FaPN, yielding an overall improvement of 1.2 - 2.6 points in AP / mIoU over FPN when paired with Faster / Mask R-CNN. In particular, our FaPN achieves the state-of-the-art of 56.7% mIoU on ADE20K when integrated within Mask-Former. The code is available from https://github.com/EMI-Group/FaPN.

cs.CV↗

arXiv2021

RelativeNAS: Relative Neural Architecture Search via Slow-Fast Learning

Despite the remarkable successes of Convolutional Neural Networks (CNNs) in computer vision, it is time-consuming and error-prone to manually design a CNN. Among various Neural Architecture Search (NAS) methods that are motivated to automate designs of high-performance CNNs, the differentiable NAS and population-based NAS are attracting increasing interests due to their unique characters. To benefit from the merits while overcoming the deficiencies of both, this work proposes a novel NAS method, RelativeNAS. As the key to efficient search, RelativeNAS performs joint learning between fast-learners (i.e. networks with relatively higher accuracy) and slow-learners in a pairwise manner. Moreover, since RelativeNAS only requires low-fidelity performance estimation to distinguish each pair of fast-learner and slow-learner, it saves certain computation costs for training the candidate architectures. The proposed RelativeNAS brings several unique advantages: (1) it achieves state-of-the-art performance on ImageNet with top-1 error rate of 24.88%, i.e. outperforming DARTS and AmoebaNet-B by 1.82% and 1.12% respectively; (2) it spends only nine hours with a single 1080Ti GPU to obtain the discovered cells, i.e. 3.75x and 7875x faster than DARTS and AmoebaNet respectively; (3) it provides that the discovered cells obtained on CIFAR-10 can be directly transferred to object detection, semantic segmentation, and keypoint detection, yielding competitive results of 73.1% mAP on PASCAL VOC, 78.7% mIoU on Cityscapes, and 68.5% AP on MSCOCO, respectively. The implementation of RelativeNAS is available at https://github.com/EMI-Group/RelativeNAS

cs.CV↗

arXiv2021

Principled Design of Translation, Scale, and Rotation Invariant Variation Operators for Metaheuristics

In the past three decades, a large number of metaheuristics have been proposed and shown high performance in solving complex optimization problems. While most variation operators in existing metaheuristics are empirically designed, this paper aims to design new operators automatically, which are expected to be search space independent and thus exhibit robust performance on different problems. For this purpose, this work first investigates the influence of translation invariance, scale invariance, and rotation invariance on the search behavior and performance of some representative operators. Then, we deduce the generic form of translation, scale, and rotation invariant operators. Afterwards, a principled approach is proposed for the automated design of operators, which searches for high-performance operators based on the deduced generic form. The experimental results demonstrate that the operators generated by the proposed approach outperform state-of-the-art ones on a variety of problems with complex landscapes and up to 1000 decision variables.

cs.NE↗

arXiv2020

Multi-objective Neural Architecture Search with Almost No Training

In the recent past, neural architecture search (NAS) has attracted increasing attention from both academia and industries. Despite the steady stream of impressive empirical results, most existing NAS algorithms are computationally prohibitive to execute due to the costly iterations of stochastic gradient descent (SGD) training. In this work, we propose an effective alternative, dubbed Random-Weight Evaluation (RWE), to rapidly estimate the performance of network architectures. By just training the last linear classification layer, RWE reduces the computational cost of evaluating an architecture from hours to seconds. When integrated within an evolutionary multi-objective algorithm, RWE obtains a set of efficient architectures with state-of-the-art performance on CIFAR-10 with less than two hours' searching on a single GPU card. Ablation studies on rank-order correlations and transfer learning experiments to ImageNet have further validated the effectiveness of RWE.

cs.CV↗

arXiv2020

Critical couplings in topological-insulator waveguide-resonator systems observed in elastic waves

Waveguides and resonators are core components in the large-scale integration of electronics, photonics, and phononics, both in existing and future scenarios. In certain situations, there is critical coupling of the two components; i.e., no energy passes through the waveguide after the incoming wave couples into the resonator. The transmission spectral characteristics resulting from this phenomenon are highly advantageous for signal filtering, switching, multiplexing, and sensing. In the present study, adopting an elastic-wave platform, we introduce topological insulator (TI), a remarkable achievement in condensed matter physics over the past decade, into a classical waveguide-ring-resonator configuration. Along with basic similarities with classical systems, a TI system has important differences and advantages, mostly owing to the spin-momentum locked transmission states at the TI boundaries. As an example, a two-port TI waveguide resonator can fundamentally eliminate upstream reflections while completely retaining useful transmission spectral characteristics, and maximize the energy in the resonator, with possible applications being novel signal processing, gyro/sensing, lasering, energy harvesting, and intense wave-matter interactions, using phonons, photons, or even electrons. The present work further enhances the confidence of using topological protection for practical device performance and functionalities, especially considering the crucial advantage of introducing (pseudo)spins to existing conventional configurations. More in-depth research on advancing phononics/photonics, especially on-chip, is foreseen.

cond-mat.mes-hall↗

arXiv2020

Evolutionary Multiobjective Optimization Driven by Generative Adversarial Networks (GANs)

Recently, increasing works have proposed to drive evolutionary algorithms using machine learning models. Usually, the performance of such model based evolutionary algorithms is highly dependent on the training qualities of the adopted models. Since it usually requires a certain amount of data (i.e. the candidate solutions generated by the algorithms) for model training, the performance deteriorates rapidly with the increase of the problem scales, due to the curse of dimensionality. To address this issue, we propose a multi-objective evolutionary algorithm driven by the generative adversarial networks (GANs). At each generation of the proposed algorithm, the parent solutions are first classified into real and fake samples to train the GANs; then the offspring solutions are sampled by the trained GANs. Thanks to the powerful generative ability of the GANs, our proposed algorithm is capable of generating promising offspring solutions in high-dimensional decision space with limited training data. The proposed algorithm is tested on 10 benchmark problems with up to 200 decision variables. Experimental results on these test problems demonstrate the effectiveness of the proposed algorithm.

cs.NE↗

arXiv2020

Evolutionary Multi-Objective Optimization Driven by Generative Adversarial Networks

Recently, more and more works have proposed to drive evolutionary algorithms using machine learning models.Usually, the performance of such model based evolutionary algorithms is highly dependent on the training qualities of the adopted models.Since it usually requires a certain amount of data (i.e. the candidate solutions generated by the algorithms) for model training, the performance deteriorates rapidly with the increase of the problem scales, due to the curse of dimensionality.To address this issue, we propose a multi-objective evolutionary algorithm driven by the generative adversarial networks (GANs).At each generation of the proposed algorithm, the parent solutions are first classified into \emph{real} and \emph{fake} samples to train the GANs; then the offspring solutions are sampled by the trained GANs.Thanks to the powerful generative ability of the GANs, our proposed algorithm is capable of generating promising offspring solutions in high-dimensional decision space with limited training data.The proposed algorithm is tested on 10 benchmark problems with up to 200 decision variables.Experimental results on these test problems demonstrate the effectiveness of the proposed algorithm.

cs.NE↗

arXiv2019

Robust Data Preprocessing for Machine-Learning-Based Disk Failure Prediction in Cloud Production Environments

To provide proactive fault tolerance for modern cloud data centers, extensive studies have proposed machine learning (ML) approaches to predict imminent disk failures for early remedy and evaluated their approaches directly on public datasets (e.g., Backblaze SMART logs). However, in real-world production environments, the data quality is imperfect (e.g., inaccurate labeling, missing data samples, and complex failure types), thereby degrading the prediction accuracy. We present RODMAN, a robust data preprocessing pipeline that refines data samples before feeding them into ML models. We start with a large-scale trace-driven study of over three million disks from Alibaba Cloud's data centers, and motivate the practical challenges in ML-based disk failure prediction. We then design RODMAN with three data preprocessing echniques, namely failure-type filtering, spline-based data filling, and automated pre-failure backtracking, that are applicable for general ML models. Evaluation on both the Alibaba and Backblaze datasets shows that RODMAN improves the prediction accuracy compared to without data preprocessing under various settings.

cs.LG↗

arXiv2019

New $\varepsilon$-regularity criteria of suitable weak solutions of the 3D Navier-Stokes equations at one scale

In this paper, by invoking the appropriate decomposition of pressure to exploit the energy hidden in pressure, we present some new $\varepsilon$-regularity criteria for suitable weak solutions of the 3D Navier-Stokes equations at one scale: for any $p,q\in [1,\infty]$ satisfying $1\leq 2/q+3/p <2$, there exists an absolute positive constant $\varepsilon$ such that $u\in L^{\infty}(Q(1/2))$ if $$\|u\|_{L^{p,q}(Q(1))}+\|Π\|_{L^{1 }(Q(1))}<\varepsilon.$$ This is an improvement of corresponding results recently proved by Guevara and Phuc in [7, Calc. Var. 56:68, 2017]. As an application of these $\varepsilon$-regularity criteria, we improve the known upper box dimension of the possible interior singular set of suitable weak solutions of the Navier-Stokes system from $975/758(\approx1.286)$ [28] to $2400/1903 (\approx1.261)$.

math.AP↗

arXiv2018

POIReviewQA: A Semantically Enriched POI Retrieval and Question Answering Dataset

Many services that perform information retrieval for Points of Interest (POI) utilize a Lucene-based setup with spatial filtering. While this type of system is easy to implement it does not make use of semantics but relies on direct word matches between a query and reviews leading to a loss in both precision and recall. To study the challenging task of semantically enriching POIs from unstructured data in order to support open-domain search and question answering (QA), we introduce a new dataset POIReviewQA. It consists of 20k questions (e.g."is this restaurant dog friendly?") for 1022 Yelp business types. For each question we sampled 10 reviews, and annotated each sentence in the reviews whether it answers the question and what the corresponding answer is. To test a system's ability to understand the text we adopt an information retrieval evaluation by ranking all the review sentences for a question based on the likelihood that they answer this question. We build a Lucene-based baseline model, which achieves 77.0% AUC and 48.8% MAP. A sentence embedding-based model achieves 79.2% AUC and 41.8% MAP, indicating that the dataset presents a challenging problem for future research by the GIR community. The result technology can help exploit the thematic content of web documents and social media for characterisation of locations.

cs.AI↗

arXiv2018

Topological Flat Band and Parity-Time Symmetry in a Honeycomb Lattice of Coupled Resonant Optical Waveguides

Two-dimensional (2D) coupled resonant optical waveguide (CROW), exhibiting topological edge states, provides an efficient platform for designing integrated topological photonic devices. In this paper, we propose an experimentally feasible design of 2D honeycomb CROW photonic structure. The characteristic optical system possesses two-fold and three-fold Dirac points at different positions in the Brillouin zone. The effective gauge fields implemented by the intrinsic pseudo-spin-orbit interaction open up topologically nontrivial bandgaps through the Dirac points. Spatial lattice geometries allow destructive wave interference, leading to a dispersionless, nearly-flat energy band in the vicinity of the three-fold Dirac point in the telecommunication frequency regime. This nontrivial nearly-flat band yields topologically protected edge states. The pertinent physical effects brought about due to non-Hermitian gain/loss medium into the honeycomb CROW device are discussed. The generalized gain-loss lattice with parity-time symmetry decouples the gain and the loss at opposite zigzag edges, leading to purely gain or loss edge channels. Meanwhile, the gain and loss effects on the armchair boundary cancel each other, giving rise to dissipationless edge states in non-Hermitian optical systems. These characteristics underpin the fundamental importance as well as the potential applications in various optical devices such as polarizers, optical couplers, beam splitters and slow light delay lines.

physics.optics↗

arXiv2017

On the Cauchy Problem of 3D Nonhomogeneous Navier-Stokes Equations with Density-Dependent Viscosity and Vacuum

We consider the global existence and large-time asymptotic behavior of strong solutions to the Cauchy problem of the three-dimensional nonhomogeneous incompressible Navier-Stokes equations with density-dependent viscosity and vacuum. We establish some key a priori exponential decay-in-time rates of the strong solutions. Then after using these estimates, we also obtain the global existence of strong solutions in the whole three-dimensional space, provided that the initial velocity is suitably small in the $\dot H^β$-norm for some $β\in(1/2,1].$ Note that this result is proved without any smallness conditions on the initial density. Moreover, the density can contain vacuum states and even have compact support initially.

math.AP↗

arXiv2017

A Monolithic Topologically Protected Phononic Circuit

Precise control of elastic waves in modes and coherences is of great use in reinforcing nowadays elastic energy harvesting/storage, nondestructive testing, wave-mater interaction, high sensitivity sensing and information processing, etc. All these implementations are expected to have elastic transmission with lower transmission losses and higher degree of freedom in transmission path. Inspired by topological states of quantum matters, especially quantum spin Hall effects (QSHEs) providing passive solutions of unique disorder-immune surface states protected by underlying nontrivial topological invariants of the bulk, thus solving severe performance trade-offs in experimentally realizable topologically ordered states. Here, we demonstrate experimentally the first elastic analogue of QSHE, by a concise phononic crystal plate with only perforated holes. Strong elastic spin-orbit coupling is realized accompanied by the first topologically-protected phononic circuits with both robustness and negligible propagation loss overcoming many circuit- and system-level performance limits induced by scattering. This elegant approach in a monolithic substrate opens up the possibility of realizing topological materials for phonons in both static and time-dependent regimes, can be immediately applied to multifarious chip-scale devices with both topological protection and massive integration, such as on-chip elastic wave-guiding, elastic splitter, elastic resonator with high quality factor, and even (pseudo-)spin filter.

cond-mat.mes-hall↗

arXiv2015

Acoustic topological insulator and robust one-way sound transport

Discovery of novel topological orders of condensed matters is of a significant interest in both fundamental and applied physics due to the associated quantum conductance behaviors and unique symmetry-protected backscattering-immune propagation against defects, which inspired similar fantastic effects in classical waves system, leading to the revolution of the manipulation of wave propagation. To date, however, only few theoretical models were proposed to realize acoustic topological states. Here, we theoretically and experimentally demonstrate a two dimensional acoustic topological insulators with acoustic analogue of quantum spin Hall Effect. Due to the band inversion mechanism near the double Dirac cones, acoustic one-way pseudospin dependent propagating edge states, corresponding to spin-plus and spin-minus, can be observed at the interface between two graphene-like acoustic crystals. We have also experimentally verified the associated topological immunity of such one-way edge states against the different lattice defects and disorders, which can always lead to inherent propagation loss and noise. We show that this unique acoustic topological phenomenon can offer a new promising application platform for the design of novel acoustic devices, such as one-way sound isolators, acoustic mode switchers, splitters, filters etc.

cond-mat.mes-hall↗

arXiv2014

Photonic analogue of quantum spin Hall effect

Symmetry-protected photonic topological insulator exhibiting robust pseudo-spin-dependent transportation, analogous to quantum spin Hall (QSH) phases and topological insulators, are of great importance in fundamental physics. Such transportation robustness is protected by time-reversal symmetry. Since electrons (fermion) and photons (boson) obey different statistics rules and associate with different time-reversal operators (i.e., Tf and Tb, respectively), whether photonic counterpart of Kramers degeneracy is topologically protected by bosonic Tb remains unidentified. Here, we construct the degenerate gapless edge states of two photonic pseudo-spins (left/right circular polarizations) in the band gap of a two-dimensional photonic crystal with strong magneto-electric coupling. We further demonstrated that the topological edge states are in fact protected by Tf rather than commonly believed Tb and their pseudo-spin dependent transportation is robust against Tf invariant impurities, discovering for the first time the topological nature of photons. Our results will pave a way towards novel photonic topological insulators and revolutionize our understandings in topological physics of fundamental particles.

physics.optics↗

arXiv2014

Accidental degeneracy of double Dirac cones in a phononic crystal

Artificial honeycomb lattices with Dirac cone dispersion provide a macroscopic platform to study the massless Dirac quasiparticles and their novel geometric phases. In this paper, a quadruple-degenerate state is achieved at the center of Brillouin zone (BZ) in a two-dimensional honeycomb lattice phononic crystal, which is a result of accidental degeneracy of two double-degenerate states. In the vicinity of the quadruple-degenerate state, the dispersion relation is linear. Such quadruple degeneracy is analyzed by rigorous representation theory of groups. Using method, a reduced Hamiltonian is obtained to describe the linear Dirac dispersion relations of such quadruple-degenerate state, which is well consistent with the simulation results. Near such accidental degeneracy, we observe some unique wave propagating properties, such as defect insensitive propagating character and Talbot effect.

cond-mat.mes-hall↗