arXiv subjects

Yu Zhang

Publications and source records attributed to Yu Zhang.

At least 901 records · Page 50Linked to original sources

arXiv2024

LEFormer: A Hybrid CNN-Transformer Architecture for Accurate Lake Extraction from Remote Sensing Imagery

Lake extraction from remote sensing images is challenging due to the complex lake shapes and inherent data noises. Existing methods suffer from blurred segmentation boundaries and poor foreground modeling. This paper proposes a hybrid CNN-Transformer architecture, called LEFormer, for accurate lake extraction. LEFormer contains three main modules: CNN encoder, Transformer encoder, and cross-encoder fusion. The CNN encoder effectively recovers local spatial information and improves fine-scale details. Simultaneously, the Transformer encoder captures long-range dependencies between sequences of any length, allowing them to obtain global features and context information. The cross-encoder fusion module integrates the local and global features to improve mask prediction. Experimental results show that LEFormer consistently achieves state-of-the-art performance and efficiency on the Surface Water and the Qinghai-Tibet Plateau Lake datasets. Specifically, LEFormer achieves 90.86% and 97.42% mIoU on two datasets with a parameter count of 3.61M, respectively, while being 20 minor than the previous best lake extraction method. The source code is available at https://github.com/BastianChen/LEFormer.

cs.CV↗

arXiv2024

Online Test-Time Adaptation of Spatial-Temporal Traffic Flow Forecasting

Accurate spatial-temporal traffic flow forecasting is crucial in aiding traffic managers in implementing control measures and assisting drivers in selecting optimal travel routes. Traditional deep-learning based methods for traffic flow forecasting typically rely on historical data to train their models, which are then used to make predictions on future data. However, the performance of the trained model usually degrades due to the temporal drift between the historical and future data. To make the model trained on historical data better adapt to future data in a fully online manner, this paper conducts the first study of the online test-time adaptation techniques for spatial-temporal traffic flow forecasting problems. To this end, we propose an Adaptive Double Correction by Series Decomposition (ADCSD) method, which first decomposes the output of the trained model into seasonal and trend-cyclical parts and then corrects them by two separate modules during the testing phase using the latest observed data entry by entry. In the proposed ADCSD method, instead of fine-tuning the whole trained model during the testing phase, a lite network is attached after the trained model, and only the lite network is fine-tuned in the testing process each time a data entry is observed. Moreover, to satisfy that different time series variables may have different levels of temporal drift, two adaptive vectors are adopted to provide different weights for different time series variables. Extensive experiments on four real-world traffic flow forecasting datasets demonstrate the effectiveness of the proposed ADCSD method. The code is available at https://github.com/Pengxin-Guo/ADCSD.

cs.LG↗

arXiv2024

USM-SCD: Multilingual Speaker Change Detection Based on Large Pretrained Foundation Models

We introduce a multilingual speaker change detection model (USM-SCD) that can simultaneously detect speaker turns and perform ASR for 96 languages. This model is adapted from a speech foundation model trained on a large quantity of supervised and unsupervised data, demonstrating the utility of fine-tuning from a large generic foundation model for a downstream task. We analyze the performance of this multilingual speaker change detection model through a series of ablation studies. We show that the USM-SCD model can achieve more than 75% average speaker change detection F1 score across a test set that consists of data from 96 languages. On American English, the USM-SCD model can achieve an 85.8% speaker change detection F1 score across various public and internal test sets, beating the previous monolingual baseline model by 21% relative. We also show that we only need to fine-tune one-quarter of the trainable model parameters to achieve the best model performance. The USM-SCD model exhibits state-of-the-art ASR quality compared with a strong public ASR baseline, making it suitable to handle both tasks with negligible additional computational cost.

eess.AS↗

arXiv2024

Machine Learning Framework for Modeling Exciton-Polaritons in Molecular Materials

A light-matter hybrid quasiparticle, called a polariton, is formed when molecules are strongly coupled to an optical cavity. Recent experiments have shown that polariton chemistry can manipulate chemical reactions. Polariton chemistry is a collective phenomenon and its effects increase with the number of molecules in a cavity. However, simulating an ensemble of molecules in the excited state coupled to a cavity mode is theoretically and computationally challenging. Recent advances in machine learning techniques have shown promising capabilities in modeling ground state chemical systems. This work presents a general protocol to predict excited-state properties, such as energies, transition dipoles, and non-adiabatic coupling vectors with the hierarchically interacting particle neural network. Machine learning predictions are then applied to compute potential energy surfaces and electronic spectra of a prototype azomethane molecule in the collective coupling scenario. These computational tools provide a much-needed framework to model and understand many molecules' emerging excited-state polariton chemistry.

physics.chem-ph↗

arXiv2023

PCR-CG: Point Cloud Registration via Deep Explicit Color and Geometry

In this paper, we introduce PCR-CG: a novel 3D point cloud registration module explicitly embedding the color signals into the geometry representation. Different from previous methods that only use geometry representation, our module is specifically designed to effectively correlate color into geometry for the point cloud registration task. Our key contribution is a 2D-3D cross-modality learning algorithm that embeds the deep features learned from color signals to the geometry representation. With our designed 2D-3D projection module, the pixel features in a square region centered at correspondences perceived from images are effectively correlated with point clouds. In this way, the overlapped regions can be inferred not only from point cloud but also from the texture appearances. Adding color is non-trivial. We compare against a variety of baselines designed for adding color to 3D, such as exhaustively adding per-pixel features or RGB values in an implicit manner. We leverage Predator [25] as the baseline method and incorporate our proposed module onto it. To validate the effectiveness of 2D features, we ablate different 2D pre-trained networks and show a positive correlation between the pre-trained weights and the task performance. Our experimental results indicate a significant improvement of 6.5% registration recall over the baseline method on the 3DLoMatch benchmark. We additionally evaluate our approach on SOTA methods and observe consistent improvements, such as an improvement of 2.4% registration recall over GeoTransformer as well as 3.5% over CoFiNet. Our study reveals a significant advantages of correlating explicit deep color features to the point cloud in the registration task.

cs.CV↗

arXiv2023

Adaptive Rank-based Tests for High Dimensional Mean Problems

The Wilcoxon signed-rank test and the Wilcoxon-Mann-Whitney test are commonly employed in one sample and two sample mean tests for one-dimensional hypothesis problems. For high-dimensional mean test problems, we calculate the asymptotic distribution of the maximum of rank statistics for each variable and suggest a max-type test. This max-type test is then merged with a sum-type test, based on their asymptotic independence offered by stationary and strong mixing assumptions. Our numerical studies reveal that this combined test demonstrates robustness and superiority over other methods, especially for heavy-tailed distributions.

stat.ME↗

arXiv2023

On a conjecture of R. M. Murty and V. K. Murty II

Let $ω^*(n)$ be the number of primes $p$ such that $p-1$ divides $n$. Assuming the Elliott--Halberstam Conjecture, we prove a conjecture posted by M. R. Murty and V. K. Murty in 2021 which states that $$\sum_{n\leqslant x}ω^*(n)^2\sim 2\frac{ζ(2)ζ(3)}{ζ(6)}x\log x, \quad \text{as} \quad x\rightarrow \infty.$$ The above sum was first investigated by Prachar in 1955. One of the key ingredients in our argument is the application of a sieve result on estimating various certain summations involving primes in arithmetic progressions, rather than a direct use of the Brun--Titchmarsh inequality which would not be applicable for our task.

math.NT↗

arXiv2023

Quantum-Assisted Joint Caching and Power Allocation for Integrated Satellite-Terrestrial Networks

Low earth orbit (LEO) satellite network can complement terrestrial networks for achieving global wireless coverage and improving delay-sensitive Internet services. This paper proposes an integrated satellite-terrestrial network (ISTN) architecture to provide ground users with seamless and reliable content delivery services. For optimal service provisioning in this architecture, we formulate an optimization model to maximize the network throughput by jointly optimizing content delivery policy, cache placement, and transmission power allocation. The resulting optimization model is a large-scale mixed-integer nonlinear program (MINLP) that is intractable for classical computer solvers. Inspired by quantum computing techniques, we propose a hybrid quantum-classical generalized Benders' decomposition (HQCGBD) algorithm to address this challenge. Specifically, we first exploit the generalized Benders' decomposition (GBD) to decompose the problem into a master problem and a subproblem and then leverage the state-of-art quantum annealer to solve the challenging master problem.

cs.NI↗

arXiv2023

Memory-Efficient Reversible Spiking Neural Networks

Spiking neural networks (SNNs) are potential competitors to artificial neural networks (ANNs) due to their high energy-efficiency on neuromorphic hardware. However, SNNs are unfolded over simulation time steps during the training process. Thus, SNNs require much more memory than ANNs, which impedes the training of deeper SNN models. In this paper, we propose the reversible spiking neural network to reduce the memory cost of intermediate activations and membrane potentials during training. Firstly, we extend the reversible architecture along temporal dimension and propose the reversible spiking block, which can reconstruct the computational graph and recompute all intermediate variables in forward pass with a reverse process. On this basis, we adopt the state-of-the-art SNN models to the reversible variants, namely reversible spiking ResNet (RevSResNet) and reversible spiking transformer (RevSFormer). Through experiments on static and neuromorphic datasets, we demonstrate that the memory cost per image of our reversible SNNs does not increase with the network depth. On CIFAR10 and CIFAR100 datasets, our RevSResNet37 and RevSFormer-4-384 achieve comparable accuracies and consume 3.79x and 3.00x lower GPU memory per image than their counterparts with roughly identical model complexity and parameters. We believe that this work can unleash the memory constraints in SNN training and pave the way for training extremely large and deep SNNs. The code is available at https://github.com/mi804/RevSNN.git.

cs.CV↗

arXiv2023

In-phase schooling hinders linear acceleration in wavy hydrofoils paired in parallel across various wavelengths

This study examines the impact of in-phase schooling on the hydrodynamic efficiency during linear acceleration in a simplified model using two undulating NACA0012 hydrofoils arranged in phalanx formation as a minimal representation of a fish school. The research focuses on key parameters, namely Strouhal (0.2-0.7) and Reynolds (1000-2000) numbers, and explores the effect of varying fish-body wavelengths (0.5-2), reflecting natural changes during actual fish linear acceleration. Contrary to expectations, in-phase schooling did not enhance acceleration performance. Both propulsive efficiency and net thrust were found to be lower compared to solitary swimming and anti-phase schooling conditions. The study also identifies and categorizes five distinct flow structure patterns within the parameters investigated, providing insight into the fluid dynamics of schooling fish during acceleration.

physics.flu-dyn↗

arXiv2023

Unify word-level and span-level tasks: NJUNLP's Participation for the WMT2023 Quality Estimation Shared Task

We introduce the submissions of the NJUNLP team to the WMT 2023 Quality Estimation (QE) shared task. Our team submitted predictions for the English-German language pair on all two sub-tasks: (i) sentence- and word-level quality prediction; and (ii) fine-grained error span detection. This year, we further explore pseudo data methods for QE based on NJUQE framework (https://github.com/NJUNLP/njuqe). We generate pseudo MQM data using parallel data from the WMT translation task. We pre-train the XLMR large model on pseudo QE data, then fine-tune it on real QE data. At both stages, we jointly learn sentence-level scores and word-level tags. Empirically, we conduct experiments to find the key hyper-parameters that improve the performance. Technically, we propose a simple method that covert the word-level outputs to fine-grained error span results. Overall, our models achieved the best results in English-German for both word-level and fine-grained error span detection sub-tasks by a considerable margin.

cs.CL↗

arXiv2023

Transforming Visual Scene Graphs to Image Captions

We propose to Transform Scene Graphs (TSG) into more descriptive captions. In TSG, we apply multi-head attention (MHA) to design the Graph Neural Network (GNN) for embedding scene graphs. After embedding, different graph embeddings contain diverse specific knowledge for generating the words with different part-of-speech, e.g., object/attribute embedding is good for generating nouns/adjectives. Motivated by this, we design a Mixture-of-Expert (MOE)-based decoder, where each expert is built on MHA, for discriminating the graph embeddings to generate different kinds of words. Since both the encoder and decoder are built based on the MHA, as a result, we construct a homogeneous encoder-decoder unlike the previous heterogeneous ones which usually apply Fully-Connected-based GNN and LSTM-based decoder. The homogeneous architecture enables us to unify the training configuration of the whole model instead of specifying different training strategies for diverse sub-networks as in the heterogeneous pipeline, which releases the training difficulty. Extensive experiments on the MS-COCO captioning benchmark validate the effectiveness of our TSG. The code is in: https://github.com/GaryJiajia/TSG.

cs.CV↗

arXiv2023

A Unified Framework for Unsupervised Domain Adaptation based on Instance Weighting

Despite the progress made in domain adaptation, solving Unsupervised Domain Adaptation (UDA) problems with a general method under complex conditions caused by label shifts between domains remains a formidable task. In this work, we comprehensively investigate four distinct UDA settings including closed set domain adaptation, partial domain adaptation, open set domain adaptation, and universal domain adaptation, where shared common classes between source and target domains coexist alongside domain-specific private classes. The prominent challenges inherent in diverse UDA settings center around the discrimination of common/private classes and the precise measurement of domain discrepancy. To surmount these challenges effectively, we propose a novel yet effective method called Learning Instance Weighting for Unsupervised Domain Adaptation (LIWUDA), which caters to various UDA settings. Specifically, the proposed LIWUDA method constructs a weight network to assign weights to each instance based on its probability of belonging to common classes, and designs Weighted Optimal Transport (WOT) for domain alignment by leveraging instance weights. Additionally, the proposed LIWUDA method devises a Separate and Align (SA) loss to separate instances with low similarities and align instances with high similarities. To guide the learning of the weight network, Intra-domain Optimal Transport (IOT) is proposed to enforce the weights of instances in common classes to follow a uniform distribution. Through the integration of those three components, the proposed LIWUDA method demonstrates its capability to address all four UDA settings in a unified manner. Experimental evaluations conducted on three benchmark datasets substantiate the effectiveness of the proposed LIWUDA method.

cs.CV↗

arXiv2023

Probing the dipole portal to heavy neutral leptons via meson decays at the high-luminosity LHC

We consider the dipole portal to sterile neutrinos, also called heavy neutral leptons (HNLs). The dipole interaction with the photon leads to HNL production in meson decays, as well as triggers the HNL decay into an active neutrino and a photon. HNLs with masses of order of 0.01-1 GeV are naturally long-lived if the dipole coupling is sufficiently small. We perform Monte-Carlo simulations and derive the sensitivities of the proposed FASER2 and FACET long-lived particle experiments to HNLs produced via the dipole operator in meson decays at the high-luminosity LHC. Our findings show that these future detectors will be complementary to each other, as well as to existing experiments, and will be able to probe new parts of the parameter space, especially in the case of the dipole operator coupled to the tau neutrino.

hep-ph↗

arXiv2023

FlatProxy: A DPU-centric Service Mesh Architecture for Hyperscale Cloud-native Application

Service mesh is a fundamental technology for building cloud-native applications, which ensures the stable running of a large number of services by an intermediate layer that governs communication between services. However, service mesh is not well suited for high-performance scenarios. The root cause is that the current service mesh is not suitable for the evolution of cloud-native applications. On the one hand, the service mesh built on CPU cannot listen to communication bypassing the CPU. On the other hand, service mesh includes many I/O-intensive and computationally-intensive tasks that can overload CPU cores as traffic grows beyond CPU performance. Therefore, we propose a data-centric service mesh that migrates the proxy of the service mesh to the entrance of the network. Moreover, we also design the DPU-centric FlatProxy, a data-centric service mesh based on DPU. There are three advantages to the DPU-centric service mesh. Firstly, it takes over all traffic flow in and out of the node, which expands the sense scale of the service mesh from container to node. Secondly, it improves communication performance and reduces host resource usage by offloading some functions and optimizing communication. Thirdly, it minimizes performance and security issues through the physical isolation of business services and cloud infrastructure. Compared with Envoy, the current mainstream service mesh implementation, FlatProxy reduces latency by 90\% and improves throughput by 4x in Gbps and 8x in qps, and it only occupies a small amount of CPU resources.

cs.DC↗

arXiv2023

SPColor: Semantic Prior Guided Exemplar-based Image Colorization

Exemplar-based image colorization aims to colorize a target grayscale image based on a color reference image, and the key is to establish accurate pixel-level semantic correspondence between these two images. Previous methods search for correspondence across the entire reference image, and this type of global matching is easy to get mismatch. We summarize the difficulties in two aspects: (1) When the reference image only contains a part of objects related to target image, improper correspondence will be established in unrelated regions. (2) It is prone to get mismatch in regions where the shape or texture of the object is easily confused. To overcome these issues, we propose SPColor, a semantic prior guided exemplar-based image colorization framework. Different from previous methods, SPColor first coarsely classifies pixels of the reference and target images to several pseudo-classes under the guidance of semantic prior, then the correspondences are only established locally between the pixels in the same class via the newly designed semantic prior guided correspondence network. In this way, improper correspondence between different semantic classes is explicitly excluded, and the mismatch is obviously alleviated. Besides, to better reserve the color from reference, a similarity masked perceptual loss is designed. Noting that the carefully designed SPColor utilizes the semantic prior provided by an unsupervised segmentation model, which is free for additional manual semantic annotations. Experiments demonstrate that our model outperforms recent state-of-the-art methods both quantitatively and qualitatively on public dataset.

cs.CV↗

arXiv2023

Conditional Adapters: Parameter-efficient Transfer Learning with Fast Inference

We propose Conditional Adapter (CoDA), a parameter-efficient transfer learning method that also improves inference efficiency. CoDA generalizes beyond standard adapter approaches to enable a new way of balancing speed and accuracy using conditional computation. Starting with an existing dense pretrained model, CoDA adds sparse activation together with a small number of new parameters and a light-weight training phase. Our experiments demonstrate that the CoDA approach provides an unexpectedly efficient way to transfer knowledge. Across a variety of language, vision, and speech tasks, CoDA achieves a 2x to 8x inference speed-up compared to the state-of-the-art Adapter approaches with moderate to no accuracy loss and the same parameter efficiency.

cs.CL↗

arXiv2023

Privacy-Preserving Load Forecasting via Personalized Model Obfuscation

The widespread adoption of smart meters provides access to detailed and localized load consumption data, suitable for training building-level load forecasting models. To mitigate privacy concerns stemming from model-induced data leakage, federated learning (FL) has been proposed. This paper addresses the performance challenges of short-term load forecasting models trained with FL on heterogeneous data, emphasizing privacy preservation through model obfuscation. Our proposed algorithm, Privacy Preserving Federated Learning (PPFL), incorporates personalization layers for localized training at each smart meter. Additionally, we employ a differentially private mechanism to safeguard against data leakage from shared layers. Simulations on the NREL ComStock dataset corroborate the effectiveness of our approach.

cs.CR↗