arXiv subjects

Hang Zheng

Publications and source records attributed to Hang Zheng.

At least 37 records · Page 2Linked to original sources

arXiv2024

Dflow, a Python framework for constructing cloud-native AI-for-Science workflows

In the AI-for-science era, scientific computing scenarios such as concurrent learning and high-throughput computing demand a new generation of infrastructure that supports scalable computing resources and automated workflow management on both cloud and high-performance supercomputers. Here we introduce Dflow, an open-source Python toolkit designed for scientists to construct workflows with simple programming interfaces. It enables complex process control and task scheduling across a distributed, heterogeneous infrastructure, leveraging containers and Kubernetes for flexibility. Dflow is highly observable and can scale to thousands of concurrent nodes per workflow, enhancing the efficiency of complex scientific computing tasks. The basic unit in Dflow, known as an Operation (OP), is reusable and independent of the underlying infrastructure or context. Dozens of workflow projects have been developed based on Dflow, spanning a wide range of projects. We anticipate that the reusability of Dflow and its components will encourage more scientists to publish their workflows and OP components. These components, in turn, can be adapted and reused in various contexts, fostering greater collaboration and innovation in the scientific community.

cs.DC↗

arXiv2024

Physical Parameters of 11,100 Short-Period ASAS-SN Eclipsing Contact Binaries

Starting from more than 11,200 short-period (less than 0.5 days) EW-type eclipsing binary candidates with the All-Sky Automated Survey for Supernovae (ASAS-SN) V-band light curves, we use MCMC and neural networks (NNs) to obtain the mass ratio ($q$), orbital inclination ($incl$), fill-out factor ($f$) and temperature ratio ($T_s/T_p$). After cross-matching with the Gaia DR3 database, the final sample contains parameters of 2,399 A-type and 8,712 W-type contact binaries (CBs). We present the distributions of parameters of these 11,111 short-period CBs. The mass ratio ($q$) and fill-out factor ($f$) are found to obey log-normal distributions, and the remaining parameters obey normal distributions. There is a significant period-temperature correlation of these CBs. Additionally, the temperature ratio (${T_s}$/${T_p}$) tends to increase as the orbital period decreases for W-type CBs. There is no significant correlation between them for A-type CBs. The mass ratio and fill-out factor ($q-f$) diagram suggest there is no significant correlation between these two parameters. A clear correlation exists between the mass ratio and radius ratio. The radius ratio increases with the mass ratio. Moreover, the deep fill-out CBs tend to fall on the upper boundary of the $q$$-$${R_s}$/${R_p}$ distribution, while the shallow fill-out CBs fall on the lower boundary.

astro-ph.SR↗

arXiv2023

FusionFormer: A Multi-sensory Fusion in Bird's-Eye-View and Temporal Consistent Transformer for 3D Object Detection

Multi-sensor modal fusion has demonstrated strong advantages in 3D object detection tasks. However, existing methods that fuse multi-modal features require transforming features into the bird's eye view space and may lose certain information on Z-axis, thus leading to inferior performance. To this end, we propose a novel end-to-end multi-modal fusion transformer-based framework, dubbed FusionFormer, that incorporates deformable attention and residual structures within the fusion encoding module. Specifically, by developing a uniform sampling strategy, our method can easily sample from 2D image and 3D voxel features spontaneously, thus exploiting flexible adaptability and avoiding explicit transformation to the bird's eye view space during the feature concatenation process. We further implement a residual structure in our feature encoder to ensure the model's robustness in case of missing an input modality. Through extensive experiments on a popular autonomous driving benchmark dataset, nuScenes, our method achieves state-of-the-art single model performance of 72.6% mAP and 75.1% NDS in the 3D object detection task without test time augmentation.

cs.CV↗

arXiv2023

FusionAD: Multi-modality Fusion for Prediction and Planning Tasks of Autonomous Driving

Building a multi-modality multi-task neural network toward accurate and robust performance is a de-facto standard in perception task of autonomous driving. However, leveraging such data from multiple sensors to jointly optimize the prediction and planning tasks remains largely unexplored. In this paper, we present FusionAD, to the best of our knowledge, the first unified framework that fuse the information from two most critical sensors, camera and LiDAR, goes beyond perception task. Concretely, we first build a transformer based multi-modality fusion network to effectively produce fusion based features. In constrast to camera-based end-to-end method UniAD, we then establish a fusion aided modality-aware prediction and status-aware planning modules, dubbed FMSPnP that take advantages of multi-modality features. We conduct extensive experiments on commonly used benchmark nuScenes dataset, our FusionAD achieves state-of-the-art performance and surpassing baselines on average 15% on perception tasks like detection and tracking, 10% on occupancy prediction accuracy, reducing prediction error from 0.708 to 0.389 in ADE score and reduces the collision rate from 0.31% to only 0.12%.

cs.CV↗

arXiv2023

CMDFusion: Bidirectional Fusion Network with Cross-modality Knowledge Distillation for LIDAR Semantic Segmentation

2D RGB images and 3D LIDAR point clouds provide complementary knowledge for the perception system of autonomous vehicles. Several 2D and 3D fusion methods have been explored for the LIDAR semantic segmentation task, but they suffer from different problems. 2D-to-3D fusion methods require strictly paired data during inference, which may not be available in real-world scenarios, while 3D-to-2D fusion methods cannot explicitly make full use of the 2D information. Therefore, we propose a Bidirectional Fusion Network with Cross-Modality Knowledge Distillation (CMDFusion) in this work. Our method has two contributions. First, our bidirectional fusion scheme explicitly and implicitly enhances the 3D feature via 2D-to-3D fusion and 3D-to-2D fusion, respectively, which surpasses either one of the single fusion schemes. Second, we distillate the 2D knowledge from a 2D network (Camera branch) to a 3D network (2D knowledge branch) so that the 3D network can generate 2D information even for those points not in the FOV (field of view) of the camera. In this way, RGB images are not required during inference anymore since the 2D knowledge branch provides 2D information according to the 3D LIDAR input. We show that our CMDFusion achieves the best performance among all fusion-based methods on SemanticKITTI and nuScenes datasets. The code will be released at https://github.com/Jun-CEN/CMDFusion.

cs.CV↗

arXiv2023

Uni-QSAR: an Auto-ML Tool for Molecular Property Prediction

Recently deep learning based quantitative structure-activity relationship (QSAR) models has shown surpassing performance than traditional methods for property prediction tasks in drug discovery. However, most DL based QSAR models are restricted to limited labeled data to achieve better performance, and also are sensitive to model scale and hyper-parameters. In this paper, we propose Uni-QSAR, a powerful Auto-ML tool for molecule property prediction tasks. Uni-QSAR combines molecular representation learning (MRL) of 1D sequential tokens, 2D topology graphs, and 3D conformers with pretraining models to leverage rich representation from large-scale unlabeled data. Without any manual fine-tuning or model selection, Uni-QSAR outperforms SOTA in 21/22 tasks of the Therapeutic Data Commons (TDC) benchmark under designed parallel workflow, with an average performance improvement of 6.09\%. Furthermore, we demonstrate the practical usefulness of Uni-QSAR in drug discovery domains.

q-bio.BM↗

arXiv2023

Multiphoton resonance band and Bloch-Siegert shift in a bichromatically driven qubit

We study the resonance and dynamics of a qubit exposed to a strong aperiodic bichromatic field by using a periodic counter-rotating hybridized rotating wave (CHRW) Hamiltonian, which is derived from the original Hamiltonian with the unitary transformations under a reasonable approximation and enables the application of the Floquet theory. It is found that the consistency between the CHRW results and numerically exact generalized-Floquet-theory (GFT) results in the valid regime of the former while the widely used rotating-wave approximation (RWA) breaks down. We illustrate that the resonance exhibits band structure and the Bloch-Siegert shifts induced by the counter-rotating couplings of the bichromatic field become notable at the multiphoton resonance band. In addition, the CHRW method is found to have a great advantage of efficiency over the GFT approach particularly in the low beat-frequency case where the latter converges very slowly. The present CHRW method provides a highly efficient way to calculate the resonance frequency incorporating the Bloch-Siegert shift and provides insights into the effects of the counter-rotating couplings of the bichromatic field in the strong-driving regimes.

quant-ph↗

arXiv2023

Do Deep Learning Methods Really Perform Better in Molecular Conformation Generation?

Molecular conformation generation (MCG) is a fundamental and important problem in drug discovery. Many traditional methods have been developed to solve the MCG problem, such as systematic searching, model-building, random searching, distance geometry, molecular dynamics, Monte Carlo methods, etc. However, they have some limitations depending on the molecular structures. Recently, there are plenty of deep learning based MCG methods, which claim they largely outperform the traditional methods. However, to our surprise, we design a simple and cheap algorithm (parameter-free) based on the traditional methods and find it is comparable to or even outperforms deep learning based MCG methods in the widely used GEOM-QM9 and GEOM-Drugs benchmarks. In particular, our design algorithm is simply the clustering of the RDKIT-generated conformations. We hope our findings can help the community to revise the deep learning methods for MCG. The code of the proposed algorithm could be found at https://gist.github.com/ZhouGengmo/5b565f51adafcd911c0bc115b2ef027c.

cs.CE↗

arXiv2023

Do Deep Learning Models Really Outperform Traditional Approaches in Molecular Docking?

Molecular docking, given a ligand molecule and a ligand binding site (called ``pocket'') on a protein, predicting the binding mode of the protein-ligand complex, is a widely used technique in drug design. Many deep learning models have been developed for molecular docking, while most existing deep learning models perform docking on the whole protein, rather than on a given pocket as the traditional molecular docking approaches, which does not match common needs. What's more, they claim to perform better than traditional molecular docking, but the approach of comparison is not fair, since traditional methods are not designed for docking on the whole protein without a given pocket. In this paper, we design a series of experiments to examine the actual performance of these deep learning models and traditional methods. For a fair comparison, we decompose the docking on the whole protein into two steps, pocket searching and docking on a given pocket, and build pipelines to evaluate traditional methods and deep learning methods respectively. We find that deep learning models are actually good at pocket searching, but traditional methods are better than deep learning models at docking on given pockets. Overall, our work explicitly reveals some potential problems in current deep learning models for molecular docking and provides several suggestions for future works.

q-bio.BM↗

arXiv2023

3D Molecular Generation via Virtual Dynamics

Structure-based drug design, i.e., finding molecules with high affinities to the target protein pocket, is one of the most critical tasks in drug discovery. Traditional solutions, like virtual screening, require exhaustively searching on a large molecular database, which are inefficient and cannot return novel molecules beyond the database. The pocket-based 3D molecular generation model, i.e., directly generating a molecule with a 3D structure and binding position in the pocket, is a new promising way to address this issue. Herein, we propose VD-Gen, a novel pocket-based 3D molecular generation pipeline. VD-Gen consists of several carefully designed stages to generate fine-grained 3D molecules with binding positions in the pocket cavity end-to-end. Rather than directly generating or sampling atoms with 3D positions in the pocket like in early attempts, in VD-Gen, we first randomly initialize many virtual particles in the pocket; then iteratively move these virtual particles, making the distribution of virtual particles approximate the distribution of molecular atoms. After virtual particles are stabilized in 3D space, we extract a 3D molecule from them. Finally, we further refine atoms in the extracted molecule by iterative movement again, to get a high-quality 3D molecule, and predict a confidence score for it. Extensive experiment results on pocket-based molecular generation demonstrate that VD-Gen can generate novel 3D molecules to fill the target pocket cavity with high binding affinities, significantly outperforming previous baselines.

q-bio.BM↗

arXiv2023

Bifurcation of the traveling wave solutions in a perturbed $(1 + 1)$-dimensional dispersive long wave equation via a geometric approach

Choosing $κ$ (horizontal ordinate of the saddle point associated to the homoclinic orbit) as bifurcation parameter, bifurcations of the travelling wave solutions is studied in a perturbed $(1 + 1)$-dimensional dispersive long wave equation. The solitary wave solution exists at a suitable wave speed $c$ for the bifurcation parameter $κ\in (0,1-\frac{\sqrt3}{3})\cup (1+\frac{\sqrt3}{3},2)$, while the kink and anti-kink wave solutions exist at a unique wave speed $c^*=\sqrt{15}/3$ for $κ=0$ or $κ=2$. The methods are based on the geometric singular perturbation (GSP, for short) approach, Melnikov method and invariant manifolds theory. Interestingly, not only the explicit analytical expression of the complicated homoclinic Melnikov integral is directly obtained for the perturbed long wave equation, but also the explicit analytical expression of the limit wave speed is directly given. Numerical simulations are utilized to verify our mathematical results.

math.AP↗

arXiv2022

A New Period Determination Method for Periodic Variable Stars

Variable stars play a key role in understanding the Milky Way and the universe. The era of astronomical big data presents new challenges for quick identification of interesting and important variable stars. Accurately estimating the periods is the most important step to distinguish different types of variable stars. Here, we propose a new method of determining the variability periods. By combining the statistical parameters of the light curves, the colors of the variables, the window function and the GLS algorithm, the aperiodic variables are excluded and the periodic variables are divided into eclipsing binaries and NEB variables (other types of periodic variable stars other than eclipsing binaries), the periods of the two main types of variables are derived. We construct a random forest classifier based on 241,154 periodic variables from the ASAS-SN and OGLE datasets of variables. The random forest classifier is trained on 17 features, among which 11 are extracted from the light curves and 6 are from the Gaia Early DR3, ALLWISE and 2MASS catalogs. The variables are classified into 7 superclasses and 17 subclasses. In comparison with the ASAS-SN and OGLE catalogs, the classification accuracy is generally above approximately 82% and the period accuracy is 70%-99%. To further test the reliability of the new method and classifier, we compare our results with the results of Chen et al. (2020) for ZTF DR2. The classification accuracy is generally above 70%. The period accuracy of the EW and SR variables is 50% and 53%, respectively. And the period accuracy of other types of variables is 65%-98%.

astro-ph.IM↗

arXiv2022

Does solitary wave solution persist for the long wave equation with small perturbations?

In this paper, persistence of solitary wave solutions of the regularized long wave equation with small perturbations are investigated by the geometric singular perturbation theory. Two different kinds of the perturbations are considered in this paper: one is the weak backward diffusion and dissipation, the other is the Marangoni effects. Indeed, the solitary wave persists under small perturbations. Furthermore, the different perturbations do affect the proper wave speed ensuring the persistence of the solitary waves. Finally, numerical simulations are utilized to confirm the theoretical results.

math.AP↗

arXiv2022

Engineered Tunable Decay Rate and Controllable Dissipative Dynamics

We investigate the steering dissipative dynamics of a two-level system (qubit) by means of the modulation of an assisted tunneling degree of freedom which is described by a quantum-oscillator spin-boson model. Our results reveal that the decoherence rate of the qubit can be significantly suppressed and simultaneously its quality factor is enhanced. Moreover, the modulated dynamical susceptibility exhibits a multi-peak feature which is indicative of the underlying structure and measurable in experiment. Our findings demonstrate that the interplay between the combined degrees of freedom and the qubit is crucial for reducing the dissipation of qubit and expanding the coherent regime of quantum operation much large. The strategy might be used to fight against deterioration of quantum coherence in quantum information processing.

quant-ph↗

arXiv2021

Geometric phase and non-adiabatic resonance of the Rabi model

We investigate the effects of counterrotating terms on geometric phase and its relation to the resonance of the Rabi model. We apply the unitary transformation with a single parameter to the Rabi model and obtain the transformed Hamiltonian involving multiple harmonic terms. By combining the counter-rotating-hybridized rotating-wave method with time-dependent perturbation theory, we solve systematically time evolution operator and then obtain the geometric phase of the two-level system. Our results are beyond adiabatic approximation and rotating-wave approximation (RWA). Higher-order harmonic resonance happens when driving frequency is equal to higher-order subharmonic of the Rabi frequency. In comparison with numerically exact results, our calculated results are accurate over a wide range of parameters space, especially in higher-order harmonic resonance regimes. In these regimes we demonstrate geometric phases change dramatically while those of the RWA are smooth. The RWA is thoroughly invalid even if the driving strength is extremely weak. We find it is the higher-order harmonic terms that play an important role on the cyclic state and demonstrate the characters of geometric phase in higher-order harmonic resonance regime. We also present analytical formalism of the change rate of geometric phase and quasienergies, which agree well with numerically exact ones even in the strong driving case. The developed method can be applied to explore the dynamics of strongly driven qubits and physical properties of higher-order harmonic processes.

quant-ph↗

arXiv2021

Photon-Inter-Correlation Optical Communication

The development of modern technology extends human presence beyond cislunar space and onto other planets, which presents an urgent need for high-capacity, long-distance and interplanetary communication. Communication using photons as carriers has a high channel capacity, but the optical diffraction limit in deep space leads to inevitable huge geometric loss, setting an insurmountable transmission distance for existing optical communication technologies. Here, we propose and experimentally demonstrate a photon-inter-correlation optical communication (PICOC) against an ultra-high channel loss. We treat light as a stream of photons, and retrieve the additional information of internal correlation and photon statistics globally from extremely weak pulse sequences. We successfully manage to build high-fidelity communication channel with a loss up to 160dB by separating a single-photon signal embedded in a noise ten times higher. With only commercially available telescopes, PICOC allows establishment of communication links from Mars to Earth communication using a milliwatt laser, and from the edge of the solar system to Earth using a few watts laser.

astro-ph.IM↗

arXiv2020

Plateau dynamics with quantized oscillations of a strongly driven qubit

We present an interesting dynamical temporal localization of a strongly driven two-level system (TLS), a plateau with quantized oscillation, by an analytical and transparent method, the counter-rotating-hybridized rotating-wave (CHRW) method. This approach, which is based on unitary transformations with a single parameter, treats the rotating and counter-rotating terms on equal footing. In the unitarily transformed representation, we find that it is the multiple-harmonic terms shown in the transformed Hamiltonian that make a crucial contribution to the generation of the exotic plateau phenomenon. By comparing the results of the numerically exact calculation and several other methods, we show that the CHRW results obtained by analytical formalism involving the collective effects of multiple harmonics are in good agreement with the numerical results, which illustrates not only the general tendency of the dynamical evolution of strongly driven TLS, but also the interesting phenomena of plateaus. The developed CHRW method reveals two kinds of plateau patterns: zigzag plateau and armchair plateau. The plateau phenomenon has a periodical pattern whose frequency is double the driving frequency, and possesses quantized oscillations the number of which has a certain, precise value. Besides, fast oscillation is produced on every plateau which is determined by the relevant driving parameters of the TLS. Our main results are as follows: (i) in the large-amplitude oscillatory case, it turns out that the collective effects of all even harmonics contribute to the generation of zigzag plateau with quantized oscillation; (ii) in the small-amplitude oscillatory case, the dynamics of the coherent destruction of tunneling under strong driving is exactly exhibited by including the odd-harmonic effect, namely, armchair plateau possessing a two-stair structure rather than the complete destruction.

quant-ph↗

arXiv2020

Role of generalized parity in the symmetry of fluorescence spectrum from two-level systems under periodic frequency modulation

We study the origin of the symmetry of the fluorescence spectrum from the two-level system subjected to a low-frequency periodic modulation and a near-resonant high-frequency monochromatic excitation by using the analytical and numerical methods based on the Floquet theory. We find that the fundamental origin of symmetry of the spectrum can be attributed to the presence of the generalized parity of the Floquet states, which depends on the driving parameters. The absence of the generalized parity can lead to the asymmetry of the spectrum. Based on the generalized parity, the conditions for the symmetry and asymmetry of the spectrum can be derived, which succeeds in predicting symmetry and asymmetry of the spectrum for the harmonic, biharmonic, and multiharmonic modulations. Moreover, we find that the secular approximation widely used in the analytical calculation may lead to artifact symmetry of the spectrum that vanishes when such approximation is avoided. The present study provides a significant perspective on the origin of the symmetry of the spectrum.

quant-ph↗