arXiv subjects

Jiahui Zhang

Publications and source records attributed to Jiahui Zhang.

At least 73 records · Page 4Linked to original sources

arXiv2024

Resilient Growth of Highly Crystalline Topological Insulator-Superconductor Heterostructure Enabled by Ex-situ Nitride Film

Highly crystalline and easily feasible topological insulator-superconductor (TI-SC) heterostructures are crucial for the development of practical topological qubit devices. The optimal superconducting layer for TI-SC heterostructures should be highly resilient against external contaminations and structurally compatible with TIs. In this study, we provide a solution to this challenge by showcasing the growth of a highly crystalline TI-SC heterostructure using refractory TiN (111) as the superconducting layer. This approach can eliminate the need for in-situ cleaving or growth. More importantly, the TiN surface shows high resilience against contaminations during air exposure, as demonstrated by the successful recyclable growth of Bi2Se3. Our findings indicate that TI-SC heterostructures based on nitride films are compatible with device fabrication techniques, paving a path to the realization of practical topological qubit devices in the future.

cond-mat.mes-hall↗

arXiv2024

Large-scale atomistic study of plasticity in amorphous gallium oxide with a machine-learning potential

Compared to the widely investigated crystalline polymorphs of gallium oxide (Ga2O3), knowledge about its amorphous state is still limited. With the help of a machine-learning interatomic potential, we conducted large-scale atomistic simulations to investigate the glass transition and mechanical behavior of amorphous Ga2O3 (a-Ga2O3). During the quenching simulations, amorphization of gallium oxide melt is observed at ultrahigh cooling rates, including a distinct glass transition. The final densities at room temperature have up to 4% variance compared to experiments. The glass transition temperature is evaluated to range from 1234 K to 1348 K at different cooling rates. Structural analysis of the amorphous structure shows evident similarities in structural properties between a-Ga2O3 and amorphous alumina (a-Al2O3), such as radial distribution function, coordination distribution, and bond angle distribution. An amorphous gallium oxide structure that contains approximately one million atoms is prepared for the tension simulation. A highly plastic behavior is observed at room temperature in the tension simulations, comparable to amorphous alumina. With quantitative characterization methods, we show that a-Ga2O3 can possibly has a higher nucleation rate of localized plastic strain events compared to a-Al2O3, which can increase the material's resistance to shear banding formation during deformation.

cond-mat.mtrl-sci↗

arXiv2024

Quantum state engineering in a five-state chainwise system by coincident pulse technique

In this paper, an exact analytical solution is presented for achieving coherent population transfer and creating arbitrary coherent superposition states in a five-state chainwise system by a train of coincident pulses. We show that the solution of a five-state chainwise system can be reduced to an equivalent three-state $Λ$-type one with the simplest resonant coupling under the assumption of adiabatic elimination (AE) together with a requirement of the relation among the four incident pulses. In this method, all of four incident pulses at each step all have the same time dependence, but with specific magnitudes. The results show that, by using a train of appropriately coincident incident pulses, this technique enables complete population transfer, as well as the creation of arbitrary desired coherent superposition between initial and final states, while the population in all intermediate states is effectively suppressed. The complete physical explanation of the underlying mechanism is presented. The results are of potential interest in applications where high-fidelity multi-state quantum control is essential, e.g., quantum information, atom optics, formation of ultracold molecules, cavity QED, nuclear coherent population transfer, light transfer in waveguide arrays, etc.

quant-ph↗

arXiv2024

FreGS: 3D Gaussian Splatting with Progressive Frequency Regularization

3D Gaussian splatting has achieved very impressive performance in real-time novel view synthesis. However, it often suffers from over-reconstruction during Gaussian densification where high-variance image regions are covered by a few large Gaussians only, leading to blur and artifacts in the rendered images. We design a progressive frequency regularization (FreGS) technique to tackle the over-reconstruction issue within the frequency space. Specifically, FreGS performs coarse-to-fine Gaussian densification by exploiting low-to-high frequency components that can be easily extracted with low-pass and high-pass filters in the Fourier space. By minimizing the discrepancy between the frequency spectrum of the rendered image and the corresponding ground truth, it achieves high-quality Gaussian densification and alleviates the over-reconstruction of Gaussian splatting effectively. Experiments over multiple widely adopted benchmarks (e.g., Mip-NeRF360, Tanks-and-Temples and Deep Blending) show that FreGS achieves superior novel view synthesis and outperforms the state-of-the-art consistently.

cs.CV↗

arXiv2024

Formation of ultracold deeply-bound molecules via multi-state chainwise coincident pulses technique

In this paper, a theoretical method for the efficient creation and detection of deeply bound molecules in three-state $Λ$-type and five-state M-type molecular systems is proposed. The method is based on the three-state coincident pulses technique and the generalized five-state coincident pulses technique. For the three-state system, the technique can efficiently transfer the populations from the Feshbach state to the deeply-bound state via a train of $N$ pairs of resonant and coincident pump and Stokes pulses, with negligible transient populations of excited states. For the five-state system, it is found that this M-type system can be generalized into a $Λ$-type structure with the simplest resonant coupling under the assumption of large one-photon detuning together with a requirement of the relation among the four incident pulses. Thereafter, this generalized model permits us to employ the reduced three-state propagator to design four coincident pulses to achieve the desired population transfer. For the numerical study, $^{87}$Rb$_2$ is considered and, it is shown that the weakly-bound Feshbach molecules can be efficiently transferred to their deeply-bound states without strong laser pulses, and the populations of all intermediate states can be well suppressed.

physics.atom-ph↗

arXiv2024

Well-balanced path-conservative discontinuous Galerkin methods with equilibrium preserving space for two-layer shallow water equations

This paper introduces well-balanced path-conservative discontinuous Galerkin (DG) methods for two-layer shallow water equations, ensuring exactness for both still water and moving water equilibrium steady states. The approach involves approximating the equilibrium variables within the DG piecewise polynomial space, while expressing the DG scheme in the form of path-conservative schemes. To robustly handle the nonconservative products governing momentum exchange between the layers, we incorporate the theory of Dal Maso, LeFloch, and Murat (DLM) within the DG method. Additionally, linear segment paths connecting the equilibrium functions are chosen to guarantee the well-balanced property of the resulting scheme. The simple ``lake-at-rest" steady state is naturally satisfied without any modification, while a specialized treatment of the numerical flux is crucial for preserving the moving water steady state. Extensive numerical examples in one and two dimensions validate the exact equilibrium preservation of the steady state solutions and demonstrate its high-order accuracy. The performance of the method and high-resolution results further underscore its potential as a robust approach for nonconservative hyperbolic balance laws.

math.NA↗

arXiv2024

Highly Efficient Creation and Detection of Deeply-bound Molecules via Invariant-based Inverse Engineering with Feasible Modified Drivings

Stimulated Raman Adiabatic Passage (STIRAP) and its variants, such as multi-state chainwise-STIRAP allow efficiently transferring the populations in multi-state system and have been widely used to prepare ultracold deeply-bound molecules. However, their transfer efficiencies are generally imperfect. The main obstacle is the presence of losses and the requirement to make the dynamics adiabatic. To this end, in the present paper a theoretical method for the efficient and robust creation and detection of deeply-bound molecules is proposed. The simple three- and five-level systems with states chainwise coupled by optical fields are considered. In the regime of large detuning, the major molecular losses are pre-suppressed by reducing the dynamics of the three- and five-level molecular systems to those of effective two- and three-level counterparts, respectively. Consequently, two-level counterpart can be directly compatible with two kinds of "Invariant-based Inverse Engineering" (IIE) recipes, the results show that both protocols give comparable performance and have good experimental feasibility. For the five-level case, by considering a relation among the four incident pulses, we show that the M-type structure can be generalized into an effective $Lambda$-type one with the simplest resonant coupling. Therefore, this generalized model can also be directly compatible with "IIE" recipe. Numerical calculations show that the weakly-bound molecules can be efficiently transferred to their deeply-bound states without strong laser intensity, and the stability against parameter variations is well preserved. Finally, the detection of ultracold deeply-bound molecules is discussed, the results show that all the protocols allow efficient detection of molecules.

quant-ph↗

arXiv2024

Equilibrium preserving space in discontinuous Galerkin methods for hyperbolic balance laws

In this paper, we develop a general framework for the design of the arbitrary high-order well-balanced discontinuous Galerkin (DG) method for hyperbolic balance laws, including the compressible Euler equations with gravitation and the shallow water equations with horizontal temperature gradients (referred to as the Ripa model). Not only the hydrostatic equilibrium including the more complicated isobaric steady state in Ripa system, but our scheme is also well-balanced for the exact preservation of the moving equilibrium state. The strategy adopted is to approximate the equilibrium variables in the DG piecewise polynomial space, rather than the conservative variables, which is pivotal in the well-balanced property. Our approach provides flexibility in combination with any consistent numerical flux, and it is free of the reference equilibrium state recovery and the special source term treatment. This approach enables the construction of a well-balanced method for non-hydrostatic equilibria in Euler systems. Extensive numerical examples such as moving or isobaric equilibria validate the high order accuracy and exact equilibrium preservation for various flows given by hyperbolic balance laws. With a relatively coarse mesh, it is also possible to capture small perturbations at or close to steady flow without numerical oscillations.

math.NA↗

arXiv2024

SPRINT: Scalable Policy Pre-Training via Language Instruction Relabeling

Pre-training robot policies with a rich set of skills can substantially accelerate the learning of downstream tasks. Prior works have defined pre-training tasks via natural language instructions, but doing so requires tedious human annotation of hundreds of thousands of instructions. Thus, we propose SPRINT, a scalable offline policy pre-training approach which substantially reduces the human effort needed for pre-training a diverse set of skills. Our method uses two core ideas to automatically expand a base set of pre-training tasks: instruction relabeling via large language models and cross-trajectory skill chaining through offline reinforcement learning. As a result, SPRINT pre-training equips robots with a much richer repertoire of skills. Experimental results in a household simulator and on a real robot kitchen manipulation task show that SPRINT leads to substantially faster learning of new long-horizon tasks than previous pre-training approaches. Website at https://clvrai.com/sprint.

cs.RO↗

arXiv2024

Weakly Supervised 3D Open-vocabulary Segmentation

Open-vocabulary segmentation of 3D scenes is a fundamental function of human perception and thus a crucial objective in computer vision research. However, this task is heavily impeded by the lack of large-scale and diverse 3D open-vocabulary segmentation datasets for training robust and generalizable models. Distilling knowledge from pre-trained 2D open-vocabulary segmentation models helps but it compromises the open-vocabulary feature as the 2D models are mostly finetuned with close-vocabulary datasets. We tackle the challenges in 3D open-vocabulary segmentation by exploiting pre-trained foundation models CLIP and DINO in a weakly supervised manner. Specifically, given only the open-vocabulary text descriptions of the objects in a scene, we distill the open-vocabulary multimodal knowledge and object reasoning capability of CLIP and DINO into a neural radiance field (NeRF), which effectively lifts 2D features into view-consistent 3D segmentation. A notable aspect of our approach is that it does not require any manual segmentation annotations for either the foundation models or the distillation process. Extensive experiments show that our method even outperforms fully supervised models trained with segmentation annotations in certain scenes, suggesting that 3D open-vocabulary segmentation can be effectively learned from 2D images and text-image pairs. Code is available at \url{https://github.com/Kunhao-Liu/3D-OVS}.

cs.CV↗

arXiv2024

Highly Efficient Creation and Detection of Ultracold Deeply-Bound Molecules via Chainwise Stimulated Raman Shortcut-to-Adiabatic Passage

Chainwise stimulated Raman adiabatic passage (C-STIRAP) in M-type molecular system is a good alternative in creating ultracold deeply-bound molecules when the typical STIRAP in Λ-type system does not work due to weak Frank-Condon factors between states. However, its creation efficiency under the smooth evolution is generally low. During the process, the population in the intermediate states may decay out quickly and the strong laser pulses may induce multi-photon processes. In this paper, we find that shortcut-to-adiabatic (STA) passage fits very well in improving the performance of the C-STIRAP. Currently, related discussions on the so-called chainwise stimulated Raman shortcut-to-adiabatic passage (C-STIRSAP) are rare. Here, we investigate this topic under the adiabatic elimination. Given a relation among the four incident pulses, it is quite interesting that the M-type system can be generalized into an effective Λ-type structure with the simplest resonant coupling. Consequently, all possible methods of STA for three-state system can be borrowed. We take the counter-diabatic driving and "chosen path" method as instances to demonstrate our treatment on the molecular system. Although the "chosen path" method does not work well in real three-state system if there is strong decay in the excited state, our C-STIRSAP protocol under both the two methods can create ultracold deeply-bound molecules with high efficiency in the M-type system. The evolution time is shortened without strong laser pulses and the robustness of STA is well preserved. Finally, the detection of ultracold deeply-bound molecules is discussed.

quant-ph↗

arXiv2023

WaveNeRF: Wavelet-based Generalizable Neural Radiance Fields

Neural Radiance Field (NeRF) has shown impressive performance in novel view synthesis via implicit scene representation. However, it usually suffers from poor scalability as requiring densely sampled images for each new scene. Several studies have attempted to mitigate this problem by integrating Multi-View Stereo (MVS) technique into NeRF while they still entail a cumbersome fine-tuning process for new scenes. Notably, the rendering quality will drop severely without this fine-tuning process and the errors mainly appear around the high-frequency features. In the light of this observation, we design WaveNeRF, which integrates wavelet frequency decomposition into MVS and NeRF to achieve generalizable yet high-quality synthesis without any per-scene optimization. To preserve high-frequency information when generating 3D feature volumes, WaveNeRF builds Multi-View Stereo in the Wavelet domain by integrating the discrete wavelet transform into the classical cascade MVS, which disentangles high-frequency information explicitly. With that, disentangled frequency features can be injected into classic NeRF via a novel hybrid neural renderer to yield faithful high-frequency details, and an intuitive frequency-guided sampling strategy can be designed to suppress artifacts around high-frequency regions. Extensive experiments over three widely studied benchmarks show that WaveNeRF achieves superior generalizable radiance field modeling when only given three images as input.

cs.CV↗

arXiv2023

Tailoring the mechanical properties of 3D microstructures: a deep learning and genetic algorithm inverse optimization framework

Materials-by-design has been historically challenging due to complex process-microstructure-property relations. Conventional analytical or simulation-based approaches suffer from low accuracy or long computational time and poor transferability, further limiting their applications in solving the inverse material design problem. Here, we establish a deep learning and genetic algorithm framework that integrates forward prediction and inverse exploration. This framework provides an end-to-end solution to achieve application-specific mechanical properties by microstructure optimization. In this study, we select the widely used Ti-6Al-4V to demonstrate the effectiveness of this framework by tailoring its microstructure and achieving various yield strength and elastic modulus across a large design space, while minimizing the stress concentration factor. Compared with conventional methods, our framework is efficient, versatile, and readily transferrable to other materials and properties. Paired with additive manufacturing's potential in controlling local microstructural features, our method has far-reaching potential for accelerating the development of application-specific, high-performing materials.

cond-mat.mtrl-sci↗

arXiv2023

Bootstrap Your Own Skills: Learning to Solve New Tasks with Large Language Model Guidance

We propose BOSS, an approach that automatically learns to solve new long-horizon, complex, and meaningful tasks by growing a learned skill library with minimal supervision. Prior work in reinforcement learning require expert supervision, in the form of demonstrations or rich reward functions, to learn long-horizon tasks. Instead, our approach BOSS (BOotStrapping your own Skills) learns to accomplish new tasks by performing "skill bootstrapping," where an agent with a set of primitive skills interacts with the environment to practice new skills without receiving reward feedback for tasks outside of the initial skill set. This bootstrapping phase is guided by large language models (LLMs) that inform the agent of meaningful skills to chain together. Through this process, BOSS builds a wide range of complex and useful behaviors from a basic set of primitive skills. We demonstrate through experiments in realistic household environments that agents trained with our LLM-guided bootstrapping procedure outperform those trained with naive bootstrapping as well as prior unsupervised skill acquisition methods on zero-shot execution of unseen, long-horizon tasks in new environments. Website at clvrai.com/boss.

cs.RO↗

arXiv2023

Regularized Vector Quantization for Tokenized Image Synthesis

Quantizing images into discrete representations has been a fundamental problem in unified generative modeling. Predominant approaches learn the discrete representation either in a deterministic manner by selecting the best-matching token or in a stochastic manner by sampling from a predicted distribution. However, deterministic quantization suffers from severe codebook collapse and misalignment with inference stage while stochastic quantization suffers from low codebook utilization and perturbed reconstruction objective. This paper presents a regularized vector quantization framework that allows to mitigate above issues effectively by applying regularization from two perspectives. The first is a prior distribution regularization which measures the discrepancy between a prior token distribution and the predicted token distribution to avoid codebook collapse and low codebook utilization. The second is a stochastic mask regularization that introduces stochasticity during quantization to strike a good balance between inference stage misalignment and unperturbed reconstruction objective. In addition, we design a probabilistic contrastive loss which serves as a calibrated metric to further mitigate the perturbed reconstruction objective. Extensive experiments show that the proposed quantization framework outperforms prevailing vector quantization methods consistently across different generative models including auto-regressive models and diffusion models.

cs.CV↗

arXiv2023

Pose-Free Neural Radiance Fields via Implicit Pose Regularization

Pose-free neural radiance fields (NeRF) aim to train NeRF with unposed multi-view images and it has achieved very impressive success in recent years. Most existing works share the pipeline of training a coarse pose estimator with rendered images at first, followed by a joint optimization of estimated poses and neural radiance field. However, as the pose estimator is trained with only rendered images, the pose estimation is usually biased or inaccurate for real images due to the domain gap between real images and rendered images, leading to poor robustness for the pose estimation of real images and further local minima in joint optimization. We design IR-NeRF, an innovative pose-free NeRF that introduces implicit pose regularization to refine pose estimator with unposed real images and improve the robustness of the pose estimation for real images. With a collection of 2D images of a specific scene, IR-NeRF constructs a scene codebook that stores scene features and captures the scene-specific pose distribution implicitly as priors. Thus, the robustness of pose estimation can be promoted with the scene priors according to the rationale that a 2D real image can be well reconstructed from the scene codebook only when its estimated pose lies within the pose distribution. Extensive experiments show that IR-NeRF achieves superior novel view synthesis and outperforms the state-of-the-art consistently across multiple synthetic and real datasets.

cs.CV↗

arXiv2023

Multimodal Image Synthesis and Editing: The Generative AI Era

As information exists in various modalities in real world, effective interaction and fusion among multimodal information plays a key role for the creation and perception of multimodal data in computer vision and deep learning research. With superb power in modeling the interaction among multimodal information, multimodal image synthesis and editing has become a hot research topic in recent years. Instead of providing explicit guidance for network training, multimodal guidance offers intuitive and flexible means for image synthesis and editing. On the other hand, this field is also facing several challenges in alignment of multimodal features, synthesis of high-resolution images, faithful evaluation metrics, etc. In this survey, we comprehensively contextualize the advance of the recent multimodal image synthesis and editing and formulate taxonomies according to data modalities and model types. We start with an introduction to different guidance modalities in image synthesis and editing, and then describe multimodal image synthesis and editing approaches extensively according to their model types. After that, we describe benchmark datasets and evaluation metrics as well as corresponding experimental results. Finally, we provide insights about the current research challenges and possible directions for future research. A project associated with this survey is available at https://github.com/fnzhan/Generative-AI.

cs.CV↗

arXiv2023

VMRF: View Matching Neural Radiance Fields

Neural Radiance Fields (NeRF) have demonstrated very impressive performance in novel view synthesis via implicitly modelling 3D representations from multi-view 2D images. However, most existing studies train NeRF models with either reasonable camera pose initialization or manually-crafted camera pose distributions which are often unavailable or hard to acquire in various real-world data. We design VMRF, an innovative view matching NeRF that enables effective NeRF training without requiring prior knowledge in camera poses or camera pose distributions. VMRF introduces a view matching scheme, which exploits unbalanced optimal transport to produce a feature transport plan for mapping a rendered image with randomly initialized camera pose to the corresponding real image. With the feature transport plan as the guidance, a novel pose calibration technique is designed which rectifies the initially randomized camera poses by predicting relative pose transformations between the pair of rendered and real images. Extensive experiments over a number of synthetic and real datasets show that the proposed VMRF outperforms the state-of-the-art qualitatively and quantitatively by large margins.

cs.CV↗