Searcharxiv⌕ Search

arXiv subjects

Shiwei Zhang

Publications and source records attributed to Shiwei Zhang.

At least 109 records · Page 6Linked to original sources
arXiv2023

Computation of forces and stresses in solids: Towards accurate structural optimization with auxiliary-field quantum Monte Carlo

The accurate computation of forces and other energy derivatives has been a long-standing challenge for quantum Monte Carlo methods. A number of technical obstacles contribute to this challenge. We discuss how these obstacles can be removed with the auxiliary-field quantum Monte Carlo (AFQMC) approach. AFQMC is a general, high-accuracy, many-body total-energy method for molecules and solids. The implementation of back-propagation for pure estimators allows direct calculation of gradients of the energy via the Hellmann-Feynman theorem. A planewave basis with norm-conserving pseudopotentials is used for the study of periodic bulk materials. Completeness of the planewave basis minimizes the effect of so-called Pulay terms. The ionic pseudopotentials, which can be incorporated in AFQMC in exactly the same manner as in standard independent-electron methods, regulate the force and stress estimators and eliminate any potential divergence of the Monte Carlo variances. The resulting approach allows applications of full geometry optimizations in bulk materials. It also paves the way for many-body computations of the phonon spectrum in solids.

cond-mat.mtrl-sci↗
arXiv2023

Interfacing branching random walks with Metropolis sampling: constraint release in auxiliary-field quantum Monte Carlo

We present an approach to interface branching random walks with Markov chain Monte Carlo sampling, and to switch seamlessly between the two. The approach is discussed in the context of auxiliary-field quantum Monte Carlo (AFQMC) but is applicable to other Monte Carlo calculations or simulations. In AFQMC, the formulation of branching random walks along imaginary-time is needed to realize a constraint to control the sign or phase problem. The constraint is derived from an exact gauge condition, and is in practice implemented approximately with a trial wave function or trial density matrix, which can break exactness in the algorithm. We use the generalized Metropolis algorithm to sample a selected portion of the imaginary-time path after it has been produced by the branching random walk. This interfacing allows a constraint release to follow seamlessly from the constrained-path sampling, which can reduce the systematic error from the latter. It also provides a way to improve the computation of correlation functions and observables that do not commute with the Hamiltonian. We illustrate the method in atoms and molecules, where improvements in accuracy can be clearly quantified and near-exact results are obtained. We also discuss the computation of the variance of the Hamiltonian and propose a convenient way to evaluate it stochastically without changing the scaling of AFQMC.

cond-mat.str-el↗
arXiv2023

Accurate Kohn-Sham auxiliary system from the ground state density of solids

The Kohn-Sham (KS) system is an auxiliary system whose effective potential is unknown in most cases. It is in principle determined by the ground state density, and it has been found numerically for some low-dimensional systems by inverting the KS equations starting from a given accurate density. For solids, only approximate results are available. In this work, we determine accurate exchange correlation (xc) potentials for Si and NaCl using the ground state densities obtained from Auxiliary Field Quantum Monte Carlo calculations. We show that these xc potentials can be rationalized as an ensemble of environment-adapted functions of the local density. The KS band structure can be obtained with high accuracy. The true KS band gap turns out to be larger than the prediction of the local density approximation, but significantly smaller than the measurable photoemission gap, which confirms previous estimates. Finally, our findings show that the conjecture that very different xc potentials can lead to very similar densities and other KS observables is true also in solids, which questions the meaning of details of the potentials and, at the same time, confirms the stability of the KS system.

physics.comp-ph↗
arXiv2023

Crowd Counting with Sparse Annotation

This paper presents a new annotation method called Sparse Annotation (SA) for crowd counting, which reduces human labeling efforts by sparsely labeling individuals in an image. We argue that sparse labeling can reduce the redundancy of full annotation and capture more diverse information from distant individuals that is not fully captured by Partial Annotation methods. Besides, we propose a point-based Progressive Point Matching network (PPM) to better explore the crowd from the whole image with sparse annotation, which includes a Proposal Matching Network (PMN) and a Performance Restoration Network (PRN). The PMN generates pseudo-point samples using a basic point classifier, while the PRN refines the point classifier with the pseudo points to maximize performance. Our experimental results show that PPM outperforms previous semi-supervised crowd counting methods with the same amount of annotation by a large margin and achieves competitive performance with state-of-the-art fully-supervised methods.

cs.CV↗
arXiv2023

MoLo: Motion-augmented Long-short Contrastive Learning for Few-shot Action Recognition

Current state-of-the-art approaches for few-shot action recognition achieve promising performance by conducting frame-level matching on learned visual features. However, they generally suffer from two limitations: i) the matching procedure between local frames tends to be inaccurate due to the lack of guidance to force long-range temporal perception; ii) explicit motion learning is usually ignored, leading to partial information loss. To address these issues, we develop a Motion-augmented Long-short Contrastive Learning (MoLo) method that contains two crucial components, including a long-short contrastive objective and a motion autodecoder. Specifically, the long-short contrastive objective is to endow local frame features with long-form temporal awareness by maximizing their agreement with the global token of videos belonging to the same class. The motion autodecoder is a lightweight architecture to reconstruct pixel motions from the differential features, which explicitly embeds the network with motion dynamics. By this means, MoLo can simultaneously learn long-range temporal context and motion cues for comprehensive few-shot matching. To demonstrate the effectiveness, we evaluate MoLo on five standard benchmarks, and the results show that MoLo favorably outperforms recent advanced methods. The source code is available at https://github.com/alibaba-mmai-research/MoLo.

cs.CV↗
arXiv2023

Enlarging Instance-specific and Class-specific Information for Open-set Action Recognition

Open-set action recognition is to reject unknown human action cases which are out of the distribution of the training set. Existing methods mainly focus on learning better uncertainty scores but dismiss the importance of feature representations. We find that features with richer semantic diversity can significantly improve the open-set performance under the same uncertainty scores. In this paper, we begin with analyzing the feature representation behavior in the open-set action recognition (OSAR) problem based on the information bottleneck (IB) theory, and propose to enlarge the instance-specific (IS) and class-specific (CS) information contained in the feature for better performance. To this end, a novel Prototypical Similarity Learning (PSL) framework is proposed to keep the instance variance within the same class to retain more IS information. Besides, we notice that unknown samples sharing similar appearances to known samples are easily misclassified as known classes. To alleviate this issue, video shuffling is further introduced in our PSL to learn distinct temporal information between original and shuffled samples, which we find enlarges the CS information. Extensive experiments demonstrate that the proposed PSL can significantly boost both the open-set and closed-set performance and achieves state-of-the-art results on multiple benchmarks. Code is available at https://github.com/Jun-CEN/PSL.

cs.CV↗
arXiv2023

Non-perturbative Many-Body Treatment of Molecular Magnets

Molecular magnets have received significant attention because of their potential applications in quantum information and quantum computing. A delicate balance of electron correlation, spin-orbit coupling (SOC), ligand field splitting, and other effects produces a persistent magnetic moment within each molecular magnet unit. The discovery and design of molecular magnets with improved functionalities would be greatly aided by accurate computations. However, the competition among the different effects poses a challenge for theoretical treatments. Electron correlation plays a central role, since d-, or f-element ions, which provide the magnetic states in molecular magnets, often require explicit many-body treatments. SOC, which expands the dimensionality of the Hilbert space, can also lead to non-perturbative effects in the presence of strong interaction. Furthermore, molecular magnets are large, with tens of atoms in even the smallest systems. We show how an $\textit{ab initio}$ treatment of molecular magnets can be achieved with auxiliary-field quantum Monte Carlo (AFQMC), in which electron correlation, SOC, and material specificity are included accurately and on an equal footing. The approach is demonstrated by an application to compute the zero-field splitting of a locally-linear Co$^{2+}$ complex.

cond-mat.str-el↗
arXiv2023

Coexistence of superconductivity with partially filled stripes in the Hubbard model

Combining the complementary capabilities of two of the most powerful modern computational methods, we find superconductivity in both the electron- and hole-doped regimes of the two-dimensional Hubbard model (with next nearest neighbor hopping). In the electron-doped regime, superconductivity is weaker and is accompanied by antiferromagnetic Néel correlations at low doping. The strong superconductivity on the hole-doped side coexists with stripe order, which persists into the overdoped region with weaker hole density modulation. These stripe orders, neither filled as in the pure Hubbard model (no next nearest neighbor hopping) nor half-filled as seen in previous state-of-the-art calculations, vary in fillings between 0.6 and 0.8. The resolution of the tiny energy scales separating competing orders requires exceedingly high accuracy combined with averaging and extrapolating with a wide range of system sizes and boundary conditions. These results validate the applicability of this iconic model for describing cuprate high-$T_c$ superconductivity.

cond-mat.supr-con↗
arXiv2023

CLIP-guided Prototype Modulating for Few-shot Action Recognition

Learning from large-scale contrastive language-image pre-training like CLIP has shown remarkable success in a wide range of downstream tasks recently, but it is still under-explored on the challenging few-shot action recognition (FSAR) task. In this work, we aim to transfer the powerful multimodal knowledge of CLIP to alleviate the inaccurate prototype estimation issue due to data scarcity, which is a critical problem in low-shot regimes. To this end, we present a CLIP-guided prototype modulating framework called CLIP-FSAR, which consists of two key components: a video-text contrastive objective and a prototype modulation. Specifically, the former bridges the task discrepancy between CLIP and the few-shot video task by contrasting videos and corresponding class text descriptions. The latter leverages the transferable textual concepts from CLIP to adaptively refine visual prototypes with a temporal Transformer. By this means, CLIP-FSAR can take full advantage of the rich semantic priors in CLIP to obtain reliable prototypes and achieve accurate few-shot classification. Extensive experiments on five commonly used benchmarks demonstrate the effectiveness of our proposed method, and CLIP-FSAR significantly outperforms existing state-of-the-art methods under various settings. The source code and models will be publicly available at https://github.com/alibaba-mmai-research/CLIP-FSAR.

cs.CV↗
arXiv2023

Temperature Dependence of Spin and Charge Orders in the Doped Two-Dimensional Hubbard Model

Competing and intertwined orders including inhomogeneous patterns of spin and charge are observed in many correlated electron materials, such as high-temperature superconductors. Introducing a new development of the constrained-path auxiliary-field quantum Monte Carlo (AFQMC) method, we study the interplay between thermal and quantum fluctuations in the two-dimensional Hubbard model. We obtain an accurate and systematic characterization of the evolution of the spin and charge correlations as a function of temperature $T$ and how it connects to the ground state, at three representative doping levels $δ= 1/5$, $1/8$, and $1/10$. We find increasing short-range commensurate antiferromagnetic correlations as $T$ is lowered. As the correlation length grows sufficiently large, a modulated spin-density-wave (SDW) appears. At $δ= 1/5$, the SDW saturates and remains short-ranged as $T \rightarrow 0$. In contrast, at $δ= 1/8$ and $1/10$ this evolves into a ground-state stripe phase. We study the relation between spin and charge orders and find that formation of charge order appears to be driven by that of the spin order. We identify a finite-temperature phase transition below which charge ordering sets in and discuss the implications of our results for the nature of this transition.

cond-mat.str-el↗
arXiv2023

Auto-Parallelizing Large Models with Rhino: A Systematic Approach on Production AI Platform

We present Rhino, a system for accelerating tensor programs with automatic parallelization on AI platform for real production environment. It transforms a tensor program written for a single device into an equivalent distributed program that is capable of scaling up to thousands of devices with no user configuration. Rhino firstly works on a semantically independent intermediate representation of tensor programs, which facilitates its generalization to unprecedented applications. Additionally, it implements a task-oriented controller and a distributed runtime for optimal performance. Rhino explores on a complete and systematic parallelization strategy space that comprises all the paradigms commonly employed in deep learning (DL), in addition to strided partitioning and pipeline parallelism on non-linear models. Aiming to efficiently search for a near-optimal parallel execution plan, our analysis of production clusters reveals general heuristics to speed up the strategy search. On top of it, two optimization levels are designed to offer users flexible trade-offs between the search time and strategy quality. Our experiments demonstrate that Rhino can not only re-discover the expert-crafted strategies of classic, research and production DL models, but also identify novel parallelization strategies which surpass existing systems for novel models.

cs.DC↗
arXiv2023

Expediting Distributed DNN Training with Device Topology-Aware Graph Deployment

This paper presents TAG, an automatic system to derive optimized DNN training graph and its deployment onto any device topology, for expedited training in device- and topology- heterogeneous ML clusters. We novelly combine both the DNN computation graph and the device topology graph as input to a graph neural network (GNN), and join the GNN with a search-based method to quickly identify optimized distributed training strategies. To reduce communication in a heterogeneous cluster, we further explore a lossless gradient compression technique and solve a combinatorial optimization problem to automatically apply the technique for training time minimization. We evaluate TAG with various representative DNN models and device topologies, showing that it can achieve up to 4.56x training speed-up as compared to existing schemes. TAG can produce efficient deployment strategies for both unseen DNN models and unseen device topologies, without heavy fine-tuning.

cs.LG↗
arXiv2023

The Devil is in the Wrongly-classified Samples: Towards Unified Open-set Recognition

Open-set Recognition (OSR) aims to identify test samples whose classes are not seen during the training process. Recently, Unified Open-set Recognition (UOSR) has been proposed to reject not only unknown samples but also known but wrongly classified samples, which tends to be more practical in real-world applications. The UOSR draws little attention since it is proposed, but we find sometimes it is even more practical than OSR in the real world applications, as evaluation results of known but wrongly classified samples are also wrong like unknown samples. In this paper, we deeply analyze the UOSR task under different training and evaluation settings to shed light on this promising research direction. For this purpose, we first evaluate the UOSR performance of several OSR methods and show a significant finding that the UOSR performance consistently surpasses the OSR performance by a large margin for the same method. We show that the reason lies in the known but wrongly classified samples, as their uncertainty distribution is extremely close to unknown samples rather than known and correctly classified samples. Second, we analyze how the two training settings of OSR (i.e., pre-training and outlier exposure) influence the UOSR. We find although they are both beneficial for distinguishing known and correctly classified samples from unknown samples, pre-training is also helpful for identifying known but wrongly classified samples while outlier exposure is not. In addition to different training settings, we also formulate a new evaluation setting for UOSR which is called few-shot UOSR, where only one or five samples per unknown class are available during evaluation to help identify unknown samples. We propose FS-KNNS for the few-shot UOSR to achieve state-of-the-art performance under all settings.

cs.CV↗
arXiv2023

HyRSM++: Hybrid Relation Guided Temporal Set Matching for Few-shot Action Recognition

Recent attempts mainly focus on learning deep representations for each video individually under the episodic meta-learning regime and then performing temporal alignment to match query and support videos. However, they still suffer from two drawbacks: (i) learning individual features without considering the entire task may result in limited representation capability, and (ii) existing alignment strategies are sensitive to noises and misaligned instances. To handle the two limitations, we propose a novel Hybrid Relation guided temporal Set Matching (HyRSM++) approach for few-shot action recognition. The core idea of HyRSM++ is to integrate all videos within the task to learn discriminative representations and involve a robust matching technique. To be specific, HyRSM++ consists of two key components, a hybrid relation module and a temporal set matching metric. Given the basic representations from the feature extractor, the hybrid relation module is introduced to fully exploit associated relations within and cross videos in an episodic task and thus can learn task-specific embeddings. Subsequently, in the temporal set matching metric, we carry out the distance measure between query and support videos from a set matching perspective and design a Bi-MHM to improve the resilience to misaligned instances. In addition, we explicitly exploit the temporal coherence in videos to regularize the matching process. Furthermore, we extend the proposed HyRSM++ to deal with the more challenging semi-supervised few-shot action recognition and unsupervised few-shot action recognition tasks. Experimental results on multiple benchmarks demonstrate that our method achieves state-of-the-art performance under various few-shot settings. The source code is available at https://github.com/alibaba-mmai-research/HyRSMPlusPlus.

cs.CV↗
arXiv2022

A structural optimization algorithm with stochastic forces and stresses

We propose an algorithm for optimizations in which the gradients contain stochastic noise. This arises, for example, in structural optimizations when computations of forces and stresses rely on methods involving Monte Carlo sampling, such as quantum Monte Carlo or neural network states, or are performed on quantum devices which have intrinsic noise. Our proposed algorithm is based on the combination of two key ingredients: an update rule derived from the steepest descent method, and a staged scheduling of the targeted statistical error and step-size, with position averaging. We compare it with commonly applied algorithms, including some of the latest machine learning optimization methods, and show that the algorithm consistently performs efficiently and robustly under realistic conditions. Applying this algorithm, we achieve full-degree optimizations in solids using ab initio many-body computations, by auxiliary-field quantum Monte Carlo with planewaves and pseudopotentials. A new metastable structure in Si was discovered in a mixed geometry and lattice relaxing simulation. In addition to structural optimization in materials, our algorithm can potentially be useful in other problems in various fields where optimization with noisy gradients is needed.

cond-mat.mtrl-sci↗
arXiv2022

Learning a Condensed Frame for Memory-Efficient Video Class-Incremental Learning

Recent incremental learning for action recognition usually stores representative videos to mitigate catastrophic forgetting. However, only a few bulky videos can be stored due to the limited memory. To address this problem, we propose FrameMaker, a memory-efficient video class-incremental learning approach that learns to produce a condensed frame for each selected video. Specifically, FrameMaker is mainly composed of two crucial components: Frame Condensing and Instance-Specific Prompt. The former is to reduce the memory cost by preserving only one condensed frame instead of the whole video, while the latter aims to compensate the lost spatio-temporal details in the Frame Condensing stage. By this means, FrameMaker enables a remarkable reduction in memory but keep enough information that can be applied to following incremental tasks. Experimental results on multiple challenging benchmarks, i.e., HMDB51, UCF101 and Something-Something V2, demonstrate that FrameMaker can achieve better performance to recent advanced methods while consuming only 20% memory. Additionally, under the same memory consumption conditions, FrameMaker significantly outperforms existing state-of-the-arts by a convincing margin.

cs.CV↗
arXiv2022

Optimizing DNN Compilation for Distributed Training with Joint OP and Tensor Fusion

This paper proposes DisCo, an automatic deep learning compilation module for data-parallel distributed training. Unlike most deep learning compilers that focus on training or inference on a single device, DisCo optimizes a DNN model for distributed training over multiple GPU machines. Existing single-device compilation strategies do not work well in distributed training, due mainly to communication inefficiency that they incur. DisCo generates optimized, joint computation operator and communication tensor fusion strategies to enable highly efficient distributed training. A GNN-based simulator is built to effectively estimate per-iteration training time achieved by operator/tensor fusion candidates. A backtracking search algorithm is driven by the simulator, navigating efficiently in the large strategy space to identify good operator/tensor fusion strategies that minimize distributed training time. We compare DisCo with existing DL fusion schemes and show that it achieves good training speed-up close to the ideal, full computation-communication overlap case.

cs.DC↗
arXiv2022

Prompt Combines Paraphrase: Teaching Pre-trained Models to Understand Rare Biomedical Words

Prompt-based fine-tuning for pre-trained models has proven effective for many natural language processing tasks under few-shot settings in general domain. However, tuning with prompt in biomedical domain has not been investigated thoroughly. Biomedical words are often rare in general domain, but quite ubiquitous in biomedical contexts, which dramatically deteriorates the performance of pre-trained models on downstream biomedical applications even after fine-tuning, especially in low-resource scenarios. We propose a simple yet effective approach to helping models learn rare biomedical words during tuning with prompt. Experimental results show that our method can achieve up to 6% improvement in biomedical natural language inference task without any extra parameters or training steps using few-shot vanilla prompt settings.

cs.CL↗