Searcharxiv⌕ Search

arXiv subjects

Yan Wang

Publications and source records attributed to Yan Wang.

At least 685 records · Page 38Linked to original sources
arXiv2023

Bit Allocation using Optimization

In this paper, we consider the problem of bit allocation in Neural Video Compression (NVC). First, we reveal a fundamental relationship between bit allocation in NVC and Semi-Amortized Variational Inference (SAVI). Specifically, we show that SAVI with GoP (Group-of-Picture)-level likelihood is equivalent to pixel-level bit allocation with precise rate \& quality dependency model. Based on this equivalence, we establish a new paradigm of bit allocation using SAVI. Different from previous bit allocation methods, our approach requires no empirical model and is thus optimal. Moreover, as the original SAVI using gradient ascent only applies to single-level latent, we extend the SAVI to multi-level such as NVC by recursively applying back-propagating through gradient ascent. Finally, we propose a tractable approximation for practical implementation. Our method can be applied to scenarios where performance outweights encoding speed, and serves as an empirical bound on the R-D performance of bit allocation. Experimental results show that current state-of-the-art bit allocation algorithms still have a room of $\approx 0.5$ dB PSNR to improve compared with ours. Code is available at \url{https://github.com/tongdaxu/Bit-Allocation-Using-Optimization}.

cs.CV↗
arXiv2023

Multi-messenger observations of double neutron stars in Galactic disk with gravitational and radio waves

We evaluate the prospects for radio follow-up of the double neutron stars (DNSs) in the Galactic disk that could be detected through future space-borne gravitational wave (GW) detectors. We first simulate the DNS population in the Galactic disk that is accessible to space-borne GW detectors according to the merger rate from recent LIGO results. Using the inspiraling waveform for the eccentric binary, the average number of the DNSs detectable by TianQin (TQ), LISA, and TQ+LISA are 217, 368, and 429, respectively. For the joint GW detection of TQ+LISA, the forecasted parameter estimation accuracies, based on the Fisher information matrix, for the detectable sources can reach the levels of $ΔP_{\mathrm b}/P_{\mathrm b} \lesssim 10^{-6}$, $ΔΩ\lesssim 100~{\mathrm {deg}}^2$, $Δe/e \lesssim 0.3$, and $Δ\dot{P}_{\mathrm b} / \dot{P}_{\mathrm b} \lesssim 0.02$. These estimation accuracies are fitted in the form of power-law function of signal-to-noise ratio. Next, we simulate the radio pulse emission from the possible pulsars in these DNSs according to pulsar beam geometry and the empirical distributions of spin period and luminosity. For the DNSs detectable by TQ+LISA, the average number of DNSs detectable by the follow-up pulsar searches using the Parkes, FAST, SKA1, and SKA are 8, 10, 43, and 87, respectively. Depending on the radio telescope, the average distances of these GW-detectable pulsar binaries vary from 1 to 7 kpc. Considering the dominant radiometer noise and phase jitter noise, the timing accuracy of these GW-detectable pulsars can be as low as 70 ${\rm ns}$ while the most probable value is about 100 $μ{\rm s}$.

astro-ph.HE↗
arXiv2023

Privacy-preserving Adversarial Facial Features

Face recognition service providers protect face privacy by extracting compact and discriminative facial features (representations) from images, and storing the facial features for real-time recognition. However, such features can still be exploited to recover the appearance of the original face by building a reconstruction network. Although several privacy-preserving methods have been proposed, the enhancement of face privacy protection is at the expense of accuracy degradation. In this paper, we propose an adversarial features-based face privacy protection (AdvFace) approach to generate privacy-preserving adversarial features, which can disrupt the mapping from adversarial features to facial images to defend against reconstruction attacks. To this end, we design a shadow model which simulates the attackers' behavior to capture the mapping function from facial features to images and generate adversarial latent noise to disrupt the mapping. The adversarial features rather than the original features are stored in the server's database to prevent leaked features from exposing facial information. Moreover, the AdvFace requires no changes to the face recognition network and can be implemented as a privacy-enhancing plugin in deployed face recognition systems. Extensive experimental results demonstrate that AdvFace outperforms the state-of-the-art face privacy-preserving methods in defending against reconstruction attacks while maintaining face recognition accuracy.

cs.CV↗
arXiv2023

A Counterfactual Collaborative Session-based Recommender System

Most session-based recommender systems (SBRSs) focus on extracting information from the observed items in the current session of a user to predict a next item, ignoring the causes outside the session (called outer-session causes, OSCs) that influence the user's selection of items. However, these causes widely exist in the real world, and few studies have investigated their role in SBRSs. In this work, we analyze the causalities and correlations of the OSCs in SBRSs from the perspective of causal inference. We find that the OSCs are essentially the confounders in SBRSs, which leads to spurious correlations in the data used to train SBRS models. To address this problem, we propose a novel SBRS framework named COCO-SBRS (COunterfactual COllaborative Session-Based Recommender Systems) to learn the causality between OSCs and user-item interactions in SBRSs. COCO-SBRS first adopts a self-supervised approach to pre-train a recommendation model by designing pseudo-labels of causes for each user's selection of the item in data to guide the training process. Next, COCO-SBRS adopts counterfactual inference to recommend items based on the outputs of the pre-trained recommendation model considering the causalities to alleviate the data sparsity problem. As a result, COCO-SBRS can learn the causalities in data, preventing the model from learning spurious correlations. The experimental results of our extensive experiments conducted on three real-world datasets demonstrate the superiority of our proposed framework over ten representative SBRSs.

cs.IR↗
arXiv2023

Optimal Estimator Design and Properties Analysis for Interconnected Systems with Asymmetric Information Structure

This paper studies the optimal state estimation problem for interconnected systems. Each subsystem can obtain its own measurement in real time, while, the measurements transmitted between the subsystems suffer from random delay. The optimal estimator is analytically designed for minimizing the conditional error covariance. The boundedness of the expected error covariance (EEC) is analyzed. In particular, a new condition that is easy to verify is established for the boundedness of EEC. Further, the properties of EEC with respect to the delay probability are studied. We found that there exists a critical probability such that the EEC is bounded if the delay probability is below the critical probability. Also, a lower and upper bound of the critical probability is derived. Finally, the proposed results are applied to a power system, and the effectiveness of the designed methods is illustrated by simulations.

eess.SY↗
arXiv2023

SAM Fails to Segment Anything? -- SAM-Adapter: Adapting SAM in Underperformed Scenes: Camouflage, Shadow, Medical Image Segmentation, and More

The emergence of large models, also known as foundation models, has brought significant advancements to AI research. One such model is Segment Anything (SAM), which is designed for image segmentation tasks. However, as with other foundation models, our experimental findings suggest that SAM may fail or perform poorly in certain segmentation tasks, such as shadow detection and camouflaged object detection (concealed object detection). This study first paves the way for applying the large pre-trained image segmentation model SAM to these downstream tasks, even in situations where SAM performs poorly. Rather than fine-tuning the SAM network, we propose \textbf{SAM-Adapter}, which incorporates domain-specific information or visual prompts into the segmentation network by using simple yet effective adapters. By integrating task-specific knowledge with general knowledge learnt by the large model, SAM-Adapter can significantly elevate the performance of SAM in challenging tasks as shown in extensive experiments. We can even outperform task-specific network models and achieve state-of-the-art performance in the task we tested: camouflaged object detection, shadow detection. We also tested polyp segmentation (medical image segmentation) and achieves better results. We believe our work opens up opportunities for utilizing SAM in downstream tasks, with potential applications in various fields, including medical image processing, agriculture, remote sensing, and more.

cs.CV↗
arXiv2023

Bidirectional Copy-Paste for Semi-Supervised Medical Image Segmentation

In semi-supervised medical image segmentation, there exist empirical mismatch problems between labeled and unlabeled data distribution. The knowledge learned from the labeled data may be largely discarded if treating labeled and unlabeled data separately or in an inconsistent manner. We propose a straightforward method for alleviating the problem - copy-pasting labeled and unlabeled data bidirectionally, in a simple Mean Teacher architecture. The method encourages unlabeled data to learn comprehensive common semantics from the labeled data in both inward and outward directions. More importantly, the consistent learning procedure for labeled and unlabeled data can largely reduce the empirical distribution gap. In detail, we copy-paste a random crop from a labeled image (foreground) onto an unlabeled image (background) and an unlabeled image (foreground) onto a labeled image (background), respectively. The two mixed images are fed into a Student network and supervised by the mixed supervisory signals of pseudo-labels and ground-truth. We reveal that the simple mechanism of copy-pasting bidirectionally between labeled and unlabeled data is good enough and the experiments show solid gains (e.g., over 21% Dice improvement on ACDC dataset with 5% labeled data) compared with other state-of-the-arts on various semi-supervised medical image segmentation datasets. Code is available at https://github.com/DeepMed-Lab-ECNU/BCP}.

cs.CV↗
arXiv2023

Low-Rate Smartphone Videoscopy for Microsecond Luminescence Lifetime Imaging with Machine Learning

Time-resolved techniques have been widely used in time-gated and luminescence lifetime imaging. However, traditional time-resolved systems require expensive lab equipment such as high-speed excitation sources and detectors or complicated mechanical choppers to achieve high repetition rates. Here, we present a cost-effective and miniaturized smartphone lifetime imaging system integrated with a pulsed UV LED for 2D luminescence lifetime imaging using a videoscopy-based virtual chopper (V-chopper) mechanism combined with machine learning. The V-chopper method generates a series of time-delayed images between excitation pulses and smartphone gating so that the luminescence lifetime can be measured at each pixel using a relatively low acquisition frame rate (e.g., 30 fps) without the need for excitation synchronization. Europium (Eu) complex dyes with different luminescent lifetimes ranging from microseconds to seconds were used to demonstrate and evaluate the principle of V-chopper on a 3D-printed smartphone microscopy platform. A convolutional neural network (CNN) model was developed to automatically distinguish the gated images in different decay cycles with an accuracy of >99.5%. The current smartphone V-chopper system can detect lifetime down to ~75 microseconds utilizing the default phase shift between the smartphone video rate and excitation pulses and in principle can detect much shorter lifetimes by accurately programming the time delay. This V-chopper methodology has eliminated the need for the expensive and complicated instruments used in traditional time-resolved detection and can greatly expand the applications of time-resolved lifetime technologies.

physics.ins-det↗
arXiv2023

Navigating phase diagram complexity to guide robotic inorganic materials synthesis

Efficient synthesis recipes are needed both to streamline the manufacturing of complex materials and to accelerate the realization of theoretically predicted materials. Oftentimes the solid-state synthesis of multicomponent oxides is impeded by undesired byproduct phases, which can kinetically trap reactions in an incomplete non-equilibrium state. We present a thermodynamic strategy to navigate high-dimensional phase diagrams in search of precursors that circumvent low-energy competing byproducts, while maximizing the reaction energy to drive fast phase transformation kinetics. Using a robotic inorganic materials synthesis laboratory, we perform a large-scale experimental validation of our precursor selection principles. For a set of 35 target quaternary oxides with chemistries representative of intercalation battery cathodes and solid-state electrolytes, we perform 224 reactions spanning 27 elements with 28 unique precursors. Our predicted precursors frequently yield target materials with higher phase purity than when starting from traditional precursors. Robotic laboratories offer an exciting new platform for data-driven experimental science, from which we can develop new insights into materials synthesis for both robot and human chemists.

cond-mat.mtrl-sci↗
arXiv2023

CogDL: A Comprehensive Library for Graph Deep Learning

Graph neural networks (GNNs) have attracted tremendous attention from the graph learning community in recent years. It has been widely adopted in various real-world applications from diverse domains, such as social networks and biological graphs. The research and applications of graph deep learning present new challenges, including the sparse nature of graph data, complicated training of GNNs, and non-standard evaluation of graph tasks. To tackle the issues, we present CogDL, a comprehensive library for graph deep learning that allows researchers and practitioners to conduct experiments, compare methods, and build applications with ease and efficiency. In CogDL, we propose a unified design for the training and evaluation of GNN models for various graph tasks, making it unique among existing graph learning libraries. By utilizing this unified trainer, CogDL can optimize the GNN training loop with several training techniques, such as mixed precision training. Moreover, we develop efficient sparse operators for CogDL, enabling it to become the most competitive graph library for efficiency. Another important CogDL feature is its focus on ease of use with the aim of facilitating open and reproducible research of graph learning. We leverage CogDL to report and maintain benchmark results on fundamental graph tasks, which can be reproduced and directly used by the community.

cs.SI↗
arXiv2023

Evaluating Strong Idempotence of Image Codec

In this paper, we first propose the concept of strong idempotent codec based on idempotent codec. The idempotence of codec refers to the stability of codec to re-compression. Similarly, we define the strong idempotence of codec as the stability of codec to multiple quality re-compression, which is an important feature of codec in the context of cloud transcoding. We provide a detailed example of strong idempotent codec with known source distribution. Further, we formalize a testing protocol to evaluate the strong idempotence of image codec. And finally, we evaluate the strong idempotence of current image codecs, including traditional codec and recent neural codec. Experimental results show that current image codecs are not stable to multi-round re-compression with different qualities, even if they are close to idempotent.

cs.MM↗
arXiv2023

Calibration-free BEV Representation for Infrastructure Perception

Effective BEV object detection on infrastructure can greatly improve traffic scenes understanding and vehicle-toinfrastructure (V2I) cooperative perception. However, cameras installed on infrastructure have various postures, and previous BEV detection methods rely on accurate calibration, which is difficult for practical applications due to inevitable natural factors (e.g., wind and snow). In this paper, we propose a Calibration-free BEV Representation (CBR) network, which achieves 3D detection based on BEV representation without calibration parameters and additional depth supervision. Specifically, we utilize two multi-layer perceptrons for decoupling the features from perspective view to front view and birdeye view under boxes-induced foreground supervision. Then, a cross-view feature fusion module matches features from orthogonal views according to similarity and conducts BEV feature enhancement with front view features. Experimental results on DAIR-V2X demonstrate that CBR achieves acceptable performance without any camera parameters and is naturally not affected by calibration noises. We hope CBR can serve as a baseline for future research addressing practical challenges of infrastructure perception.

cs.CV↗
arXiv2023

Experience-Based Evolutionary Algorithms for Expensive Optimization

Optimization algorithms are very different from human optimizers. A human being would gain more experiences through problem-solving, which helps her/him in solving a new unseen problem. Yet an optimization algorithm never gains any experiences by solving more problems. In recent years, efforts have been made towards endowing optimization algorithms with some abilities of experience learning, which is regarded as experience-based optimization. In this paper, we argue that hard optimization problems could be tackled efficiently by making better use of experiences gained in related problems. We demonstrate our ideas in the context of expensive optimization, where we aim to find a near-optimal solution to an expensive optimization problem with as few fitness evaluations as possible. To achieve this, we propose an experience-based surrogate-assisted evolutionary algorithm (SAEA) framework to enhance the optimization efficiency of expensive problems, where experiences are gained across related expensive tasks via a novel meta-learning method. These experiences serve as the task-independent parameters of a deep kernel learning surrogate, then the solutions sampled from the target task are used to adapt task-specific parameters for the surrogate. With the help of experience learning, competitive regression-based surrogates can be initialized using only 1$d$ solutions from the target task ($d$ is the dimension of the decision space). Our experimental results on expensive multi-objective and constrained optimization problems demonstrate that experiences gained from related tasks are beneficial for the saving of evaluation budgets on the target problem.

cs.NE↗
arXiv2023

MagicNet: Semi-Supervised Multi-Organ Segmentation via Magic-Cube Partition and Recovery

We propose a novel teacher-student model for semi-supervised multi-organ segmentation. In teacher-student model, data augmentation is usually adopted on unlabeled data to regularize the consistent training between teacher and student. We start from a key perspective that fixed relative locations and variable sizes of different organs can provide distribution information where a multi-organ CT scan is drawn. Thus, we treat the prior anatomy as a strong tool to guide the data augmentation and reduce the mismatch between labeled and unlabeled images for semi-supervised learning. More specifically, we propose a data augmentation strategy based on partition-and-recovery N$^3$ cubes cross- and within- labeled and unlabeled images. Our strategy encourages unlabeled images to learn organ semantics in relative locations from the labeled images (cross-branch) and enhances the learning ability for small organs (within-branch). For within-branch, we further propose to refine the quality of pseudo labels by blending the learned representations from small cubes to incorporate local attributes. Our method is termed as MagicNet, since it treats the CT volume as a magic-cube and N$^3$-cube partition-and-recovery process matches with the rule of playing a magic-cube. Extensive experiments on two public CT multi-organ datasets demonstrate the effectiveness of MagicNet, and noticeably outperforms state-of-the-art semi-supervised medical image segmentation approaches, with +7% DSC improvement on MACT dataset with 10% labeled images. Code is available at https://github.com/DeepMed-Lab-ECNU/MagicNet.

cs.CV↗
arXiv2023

VIMI: Vehicle-Infrastructure Multi-view Intermediate Fusion for Camera-based 3D Object Detection

In autonomous driving, Vehicle-Infrastructure Cooperative 3D Object Detection (VIC3D) makes use of multi-view cameras from both vehicles and traffic infrastructure, providing a global vantage point with rich semantic context of road conditions beyond a single vehicle viewpoint. Two major challenges prevail in VIC3D: 1) inherent calibration noise when fusing multi-view images, caused by time asynchrony across cameras; 2) information loss when projecting 2D features into 3D space. To address these issues, We propose a novel 3D object detection framework, Vehicles-Infrastructure Multi-view Intermediate fusion (VIMI). First, to fully exploit the holistic perspectives from both vehicles and infrastructure, we propose a Multi-scale Cross Attention (MCA) module that fuses infrastructure and vehicle features on selective multi-scales to correct the calibration noise introduced by camera asynchrony. Then, we design a Camera-aware Channel Masking (CCM) module that uses camera parameters as priors to augment the fused features. We further introduce a Feature Compression (FC) module with channel and spatial compression blocks to reduce the size of transmitted features for enhanced efficiency. Experiments show that VIMI achieves 15.61% overall AP_3D and 21.44% AP_BEV on the new VIC3D dataset, DAIR-V2X-C, significantly outperforming state-of-the-art early fusion and late fusion methods with comparable transmission cost.

cs.CV↗
arXiv2023

Mapping the metal-insulator phase diagram by algebraically fast-forwarding dynamics on a cloud quantum computer

Dynamical mean-field theory (DMFT) maps the local Green's function of the Hubbard model to that of the Anderson impurity model and thus gives an approximate solution of the Hubbard model from the solution of simpler quantum impurity model. Quantum and hybrid quantum-classical algorithms have been proposed to efficiently solve impurity models by preparing and evolving the ground state under the impurity Hamiltonian on a quantum computer that is assumed to have the scalability and accuracy far beyond the current state-of-the-art quantum hardware. As a proof of principle demonstration targeting the Anderson impurity model we close the DMFT loop with current noisy hardware. With a highly optimized fast-forwarding quantum circuit and a noise resilient spectral analysis we observe a Mott phase transition. Based on a Cartan decomposition, our algorithm gives a fixed depth, fast-forwarding, quantum circuit that can evolve the initial state over arbitrarily long times without time-discretization errors typical of other product decomposition formulas such as Trotter decomposition. By exploiting the structure of the fast-forwarding circuits we reduce the gate count (to 77 CNOTs after optimization), simulate the dynamics, and extract frequencies from the Anderson impurity model on noisy quantum hardware. We then demonstrate the Mott transition by mapping the full metal-insulator phase-diagram. Near the Mott phase transition, our method maintains accuracy where Trotter error would otherwise dominate due to the long-time evolution required to resolve quasiparticle resonance frequency extremely close to zero. This work presents the first computation of the Mott phase transition using noisy digital quantum hardware, made viable by a highly optimized computation in terms of gate depth, simulation error, and runtime on quantum hardware.

quant-ph↗
arXiv2023

Meta Architecture for Point Cloud Analysis

Recent advances in 3D point cloud analysis bring a diverse set of network architectures to the field. However, the lack of a unified framework to interpret those networks makes any systematic comparison, contrast, or analysis challenging, and practically limits healthy development of the field. In this paper, we take the initiative to explore and propose a unified framework called PointMeta, to which the popular 3D point cloud analysis approaches could fit. This brings three benefits. First, it allows us to compare different approaches in a fair manner, and use quick experiments to verify any empirical observations or assumptions summarized from the comparison. Second, the big picture brought by PointMeta enables us to think across different components, and revisit common beliefs and key design decisions made by the popular approaches. Third, based on the learnings from the previous two analyses, by doing simple tweaks on the existing approaches, we are able to derive a basic building block, termed PointMetaBase. It shows very strong performance in efficiency and effectiveness through extensive experiments on challenging benchmarks, and thus verifies the necessity and benefits of high-level interpretation, contrast, and comparison like PointMeta. In particular, PointMetaBase surpasses the previous state-of-the-art method by 0.7%/1.4/%2.1% mIoU with only 2%/11%/13% of the computation cost on the S3DIS datasets.

cs.CV↗
arXiv2023

DDPNAS: Efficient Neural Architecture Search via Dynamic Distribution Pruning

Neural Architecture Search (NAS) has demonstrated state-of-the-art performance on various computer vision tasks. Despite the superior performance achieved, the efficiency and generality of existing methods are highly valued due to their high computational complexity and low generality. In this paper, we propose an efficient and unified NAS framework termed DDPNAS via dynamic distribution pruning, facilitating a theoretical bound on accuracy and efficiency. In particular, we first sample architectures from a joint categorical distribution. Then the search space is dynamically pruned and its distribution is updated every few epochs. With the proposed efficient network generation method, we directly obtain the optimal neural architectures on given constraints, which is practical for on-device models across diverse search spaces and constraints. The architectures searched by our method achieve remarkable top-1 accuracies, 97.56 and 77.2 on CIFAR-10 and ImageNet (mobile settings), respectively, with the fastest search process, i.e., only 1.8 GPU hours on a Tesla V100. Codes for searching and network generation are available at: https://openi.pcl.ac.cn/PCL AutoML/XNAS.

cs.CV↗