arXiv subjects

Hao Zhou

Publications and source records attributed to Hao Zhou.

At least 307 records · Page 17Linked to original sources

arXiv2024

FROSTER: Frozen CLIP Is A Strong Teacher for Open-Vocabulary Action Recognition

In this paper, we introduce FROSTER, an effective framework for open-vocabulary action recognition. The CLIP model has achieved remarkable success in a range of image-based tasks, benefiting from its strong generalization capability stemming from pretaining on massive image-text pairs. However, applying CLIP directly to the open-vocabulary action recognition task is challenging due to the absence of temporal information in CLIP's pretraining. Further, fine-tuning CLIP on action recognition datasets may lead to overfitting and hinder its generalizability, resulting in unsatisfactory results when dealing with unseen actions. To address these issues, FROSTER employs a residual feature distillation approach to ensure that CLIP retains its generalization capability while effectively adapting to the action recognition task. Specifically, the residual feature distillation treats the frozen CLIP model as a teacher to maintain the generalizability exhibited by the original CLIP and supervises the feature learning for the extraction of video-specific features to bridge the gap between images and videos. Meanwhile, it uses a residual sub-network for feature distillation to reach a balance between the two distinct objectives of learning generalizable and video-specific features. We extensively evaluate FROSTER on open-vocabulary action recognition benchmarks under both base-to-novel and cross-dataset settings. FROSTER consistently achieves state-of-the-art performance on all datasets across the board. Project page: https://visual-ai.github.io/froster.

cs.CV↗

arXiv2024

Revisiting the Markov Property for Machine Translation

In this paper, we re-examine the Markov property in the context of neural machine translation. We design a Markov Autoregressive Transformer~(MAT) and undertake a comprehensive assessment of its performance across four WMT benchmarks. Our findings indicate that MAT with an order larger than 4 can generate translations with quality on par with that of conventional autoregressive transformers. In addition, counter-intuitively, we also find that the advantages of utilizing a higher-order MAT do not specifically contribute to the translation of longer sentences.

cs.CL↗

arXiv2024

Multi-fidelity physics constrained neural networks for dynamical systems

Physics-constrained neural networks are commonly employed to enhance prediction robustness compared to purely data-driven models, achieved through the inclusion of physical constraint losses during the model training process. However, one of the major challenges of physics-constrained neural networks consists of the training complexity especially for high-dimensional systems. In fact, conventional physics-constrained models rely on singular-fidelity data necessitating the assessment of physical constraints within high-dimensional fields, which introduces computational difficulties. Furthermore, due to the fixed input size of the neural networks, employing multi-fidelity training data can also be cumbersome. In this paper, we propose the Multi-Scale Physics-Constrained Neural Network (MSPCNN), which offers a novel methodology for incorporating data with different levels of fidelity into a unified latent space through a customised multi-fidelity autoencoder. Additionally, multiple decoders are concurrently trained to map latent representations of inputs into various fidelity physical spaces. As a result, during the training of predictive models, physical constraints can be evaluated within low-fidelity spaces, yielding a trade-off between training efficiency and accuracy. In addition, unlike conventional methods, MSPCNN also manages to employ multi-fidelity data to train the predictive model. We assess the performance of MSPCNN in two fluid dynamics problems, namely a two-dimensional Burgers' system and a shallow water system. Numerical results clearly demonstrate the enhancement of prediction accuracy and noise robustness when introducing physical constraints in low-fidelity fields. On the other hand, as expected, the training complexity can be significantly reduced by computing physical constraint loss in the low-fidelity field rather than the high-fidelity one.

cs.LG↗

arXiv2024

The Activity of Black Hole Imprints on the Internal Plateau and the Subsequent Sharp Decay

A stellar-mass black hole (BH) or a millisecond magnetar is believed to be born as the central engine of Gamma-ray bursts (GRBs). The presence of plateaus in the X-ray extended emission or afterglow of GRBs is widely accepted as an indicator of magnetar central engine, particularly those with a sharp decay (faster than $t^{-3}$), so-called internal plateau. However, an alternative model, by taking the evolution of the magnetic flux at the BH horizon into account, suggests that an internal plateau can also arise from a Blandford-Znajek (BZ) mechanism powered jet (hereafter referred to as the BZ jet). In this study, we propose that a precessional BZ jet would manifest a Quasi-Periodic Oscillation (QPO) signature on the internal plateau and the subsequent sharp decay. Such lightcurves cannot be readily explained by the activity of a short-lived, supermassive magnetar, thus favoring a Kerr BH as the central engine. The X-ray afterglow of GRB 050904, comprising nine flares, is characterized by a QPO-modulated plateau and sharp decay, which can be well reproduced by a precessional BZ jet model. Therefore, one potential clue for distinguishing between these two engines lies in whether QPO signature throughout the entire plateau and the subsequent sharp decay, as the magnetar scenario suggests a collapse at the end of the plateau.

astro-ph.HE↗

arXiv2024

Black holes as the source of dark energy: a stringent test with high-redshift JWST AGNs

Studies have proposed that there is evidence for cosmological coupling of black holes (BHs) with an index of $k\approx 3$; hence, BHs serve as the astrophysical source of dark energy. However, the data sample is limited for the redshifts of $\leq 2.5$. In recent years, the James Webb Space Telescope (JWST) has detected many high-redshift active galactic nuclei (AGNs) and quasars. Among the JWST NIRSpec-/NIRCam-resolved AGNs, three are determined to be in early-type host galaxies with a redshift of $z\sim 4.5--7$. However, their $M_{\star}$ and $M_{\rm BH}$ are in tension with the predicted cosmological coupling of black holes with $k = 3$ at a confidence level of $\sim 2σ$, which challenges the hypothesis that BHs serve as the origin of dark energy. Future work on high-redshift AGNs using the JWST will further assess such a hypothesis by identifying more early-type host galaxies in the higher mass range.

astro-ph.CO↗

arXiv2024

Multi-Energy Blended CBCT Spectral Imaging Using a Spectral Modulator with Flying Focal Spot (SMFFS)

Conventional cone-beam CT (CBCT) can be easily compromised by scatter and beam hardening artifacts, and the entanglement of scatter and spectral effects introduces additional complexity. In this work, we present the first attempt to develop a stationary spectral modulator with flying focal spot (SMFFS) technology as a promising, low-cost approach to accurately solving the X-ray scattering problem and physically enabling spectral imaging in a unified framework. To deal with the intertwined scatter-spectral challenge, we propose a novel scatter-decoupled material decomposition (SDMD) method for SMFFS based on a hypothesis of scatter similarity. Monte Carlo simulations of a pure-water cylinder phantom with different focal spot deflections show that focal spot deflections within a range of ~2 mm share quite similar scatter distributions overall. Numerical simulations using a clinical abdominal CT dataset demonstrate that SMFFS with SDMD method can achieve better material decomposition and CT number accuracy with less artifacts. Physics experiments on a tabletop CBCT system using a Gammex multi-energy CT phantom an anthropomorphic chest phantom, are carried out to demonstrate the feasibility of CBCT spectral imaging with SMFFS. For the chest phantom, the root mean square error (RMSE) in selected regions of interest (ROIs) of virtual monochromatic image (VMI) at 70 keV is 11.8 HU for SMFFS CB scan, and 14.5 and 437.6 HU for sequential 80/140 kVp (DKV) CB scan with and without scatter correction, respectively. Also, the non-uniformity among selected regions is 14.1 HU for SMFFS CB scan, and 59.4 and 184.0 HU for the DKV CB scan with and without a traditional scatter correction method, respectively. Our preliminary results show that SMFFS can enable spectral imaging with simultaneous scatter correction for CBCT and effectively improve its quantitative imaging performance.

physics.med-ph↗

arXiv2024

Fast KV-Switching and Dual-Layer Flat-Panel Detector Enabled Cone-Beam CT Joint Spectral Imaging

Purpose: Fast kV-switching (FKS) and dual-layer flat-panel detector (DL-FPD) technologies have been actively studied as promising dual-energy solutions for FPD-based cone-beam computed tomography (CBCT). However, CBCT spectral imaging is known to face challenges in obtaining accurate and robust material discrimination performance due to the limited energy separation. To further improve CBCT spectral imaging capability, this work aims to promote a source-detector joint spectral imaging solution which takes advantages of both FKS and DL-FPD, and to conduct a feasibility study on the first tabletop CBCT system with the joint spectral imaging capability developed. Methods: In this work, the first FKS and DL-FPD jointly enabled multi-energy tabletop CBCT system has been developed in our laboratory. To evaluate its spectral imaging performance, a set of physics experiments are conducted, where the multi-energy and head phantoms are scanned using the 80/105/130kVp switching pairs and projection data are collected using a prototype DL-FPD. To compensate for the slightly angular mismatch between the low- and high-energy projections in FKS, a dual-domain projection completion scheme is implemented. Afterwards material decomposition is carried out by using the maximum-likelihood method, followed by reconstruction of basis material and virtual monochromatic images. Results: The physics experiments confirmed the feasibility and superiority of the joint spectral imaging, whose CNR of the multi-energy phantom were boosted by an average improvement of 21.9%, 20.4% for water and 32.8%, 62.8% for iodine when compared with that of the FKS and DL-FPD in fan-beam and cone-beam experiments, respectively. Conclusions: A feasibility study of the joint spectral imaging for CBCT by utilizing both the FKS and DL-FPD was conducted, with the first tabletop CBCT system having such a capability being developed.

physics.med-ph↗

arXiv2023

Label Words are Anchors: An Information Flow Perspective for Understanding In-Context Learning

In-context learning (ICL) emerges as a promising capability of large language models (LLMs) by providing them with demonstration examples to perform diverse tasks. However, the underlying mechanism of how LLMs learn from the provided context remains under-explored. In this paper, we investigate the working mechanism of ICL through an information flow lens. Our findings reveal that label words in the demonstration examples function as anchors: (1) semantic information aggregates into label word representations during the shallow computation layers' processing; (2) the consolidated information in label words serves as a reference for LLMs' final predictions. Based on these insights, we introduce an anchor re-weighting method to improve ICL performance, a demonstration compression technique to expedite inference, and an analysis framework for diagnosing ICL errors in GPT2-XL. The promising applications of our findings again validate the uncovered ICL working mechanism and pave the way for future studies.

cs.CL↗

arXiv2023

Equivariant Flow Matching with Hybrid Probability Transport

The generation of 3D molecules requires simultaneously deciding the categorical features~(atom types) and continuous features~(atom coordinates). Deep generative models, especially Diffusion Models (DMs), have demonstrated effectiveness in generating feature-rich geometries. However, existing DMs typically suffer from unstable probability dynamics with inefficient sampling speed. In this paper, we introduce geometric flow matching, which enjoys the advantages of both equivariant modeling and stabilized probability dynamics. More specifically, we propose a hybrid probability path where the coordinates probability path is regularized by an equivariant optimal transport, and the information between different modalities is aligned. Experimentally, the proposed method could consistently achieve better performance on multiple molecule generation benchmarks with 4.75$\times$ speed up of sampling on average.

cs.LG↗

arXiv2023

Non-Autoregressive Document-Level Machine Translation

Non-autoregressive translation (NAT) models achieve comparable performance and superior speed compared to auto-regressive translation (AT) models in the context of sentence-level machine translation (MT). However, their abilities are unexplored in document-level MT, hindering their usage in real scenarios. In this paper, we conduct a comprehensive examination of typical NAT models in the context of document-level MT and further propose a simple but effective design of sentence alignment between source and target. Experiments show that NAT models achieve high acceleration on documents, and sentence alignment significantly enhances their performance. However, current NAT models still have a significant performance gap compared to their AT counterparts. Further investigation reveals that NAT models suffer more from the multi-modality and misalignment issues in the context of document-level MT, and current NAT models struggle with exploiting document context and handling discourse phenomena. We delve into these challenges and provide our code at \url{https://github.com/baoguangsheng/nat-on-doc}.

cs.CL↗

arXiv2023

Diffusion Glancing Transformer for Parallel Sequence to Sequence Learning

Previously, non-autoregressive models were widely perceived as being superior in generation efficiency but inferior in generation quality due to the difficulties of modeling multiple target modalities. To enhance the multi-modality modeling ability, we propose the diffusion glancing transformer, which employs a modality diffusion process and residual glancing sampling. The modality diffusion process is a discrete process that interpolates the multi-modal distribution along the decoding steps, and the residual glancing sampling approach guides the model to continuously learn the remaining modalities across the layers. Experimental results on various machine translation and text generation benchmarks demonstrate that DIFFGLAT achieves better generation accuracy while maintaining fast decoding speed compared with both autoregressive and non-autoregressive models.

cs.CL↗

arXiv2023

A Survey on Model-based, Heuristic, and Machine Learning Optimization Approaches in RIS-aided Wireless Networks

Reconfigurable intelligent surfaces (RISs) have received considerable attention as a key enabler for envisioned 6G networks, for the purpose of improving the network capacity, coverage, efficiency, and security with low energy consumption and low hardware cost. However, integrating RISs into the existing infrastructure greatly increases the network management complexity, especially for controlling a significant number of RIS elements. To unleash the full potential of RISs, efficient optimization approaches are of great importance. This work provides a comprehensive survey on optimization techniques for RIS-aided wireless communications, including model-based, heuristic, and machine learning (ML) algorithms. In particular, we first summarize the problem formulations in the literature with diverse objectives and constraints, e.g., sum-rate maximization, power minimization, and imperfect channel state information constraints. Then, we introduce model-based algorithms that have been used in the literature, such as alternating optimization, the majorization-minimization method, and successive convex approximation. Next, heuristic optimization is discussed, which applies heuristic rules for obtaining low-complexity solutions. Moreover, we present state-of-the-art ML algorithms and applications towards RISs, i.e., supervised and unsupervised learning, reinforcement learning, federated learning, graph learning, transfer learning, and hierarchical learning-based approaches. Model-based, heuristic, and ML approaches are compared in terms of stability, robustness, optimality and so on, providing a systematic understanding of these techniques. Finally, we highlight RIS-aided applications towards 6G networks and identify future challenges.

eess.SP↗

arXiv2023

On-Device Intelligence for 5G RAN: Knowledge Transfer and Federated Learning enabled UE-Centric Traffic Steering

Traffic steering (TS) is a promising approach to support various service requirements and enhance transmission reliability by distributing network traffic loads to appropriate base stations (BSs). In conventional cell-centric TS strategies, BSs make TS decisions for all user equipment (UEs) in a centralized manner, which focuses more on the overall performance of the whole cell, disregarding specific requirements of individual UE. The flourishing machine learning technologies and evolving UE-centric 5G network architecture have prompted the emergence of new TS technologies. In this paper, we propose a knowledge transfer and federated learning-enabled UE-centric (KT-FLUC) TS framework for highly dynamic 5G radio access networks (RAN). Specifically, first, we propose an attention-weighted group federated learning scheme. It enables intelligent UEs to make TS decisions autonomously using local models and observations, and a global model is defined to coordinate local TS decisions and share experiences among UEs. Secondly, considering the individual UE's limited computation and energy resources, a growing and pruning-based model compression method is introduced, mitigating the computation burden of UEs and reducing the communication overhead of federated learning. In addition, we propose a Q-value-based knowledge transfer method to initialize newcomer UEs, achieving a jump start for their training efficiency. Finally, the simulations show that our proposed KT-FLUC algorithm can effectively improve the service quality, achieving 65\% and 38\% lower delay and 52% and 57% higher throughput compared with cell-based TS and other UE-centric TS strategies, respectively.

cs.NI↗

arXiv2023

Distributed Attacks over Federated Reinforcement Learning-enabled Cell Sleep Control

Federated learning (FL) is particularly useful in wireless networks due to its distributed implementation and privacy-preserving features. However, as a distributed learning system, FL can be vulnerable to malicious attacks from both internal and external sources. Our work aims to investigate the attack models in a FL-enabled wireless networks. Specifically, we consider a cell sleep control scenario, and apply federated reinforcement learning to improve energy-efficiency. We design three attacks, namely free rider attacks, Byzantine data poisoning attacks and backdoor attacks. The simulation results show that the designed attacks can degrade the network performance and lead to lower energy-efficiency. Moreover, we also explore possible ways to mitigate the above attacks. We design a defense model called refined-Krum to defend against attacks by enabling a secure aggregation on the global server. The proposed refined- Krum scheme outperforms the existing Krum scheme and can effectively prevent wireless networks from malicious attacks, improving the system energy-efficiency performance.

cs.NI↗

arXiv2023

Revisiting the Domain Shift and Sample Uncertainty in Multi-source Active Domain Transfer

Active Domain Adaptation (ADA) aims to maximally boost model adaptation in a new target domain by actively selecting a limited number of target data to annotate.This setting neglects the more practical scenario where training data are collected from multiple sources. This motivates us to target a new and challenging setting of knowledge transfer that extends ADA from a single source domain to multiple source domains, termed Multi-source Active Domain Adaptation (MADA). Not surprisingly, we find that most traditional ADA methods cannot work directly in such a setting, mainly due to the excessive domain gap introduced by all the source domains and thus their uncertainty-aware sample selection can easily become miscalibrated under the multi-domain shifts. Considering this, we propose a Dynamic integrated uncertainty valuation framework(Detective) that comprehensively consider the domain shift between multi-source domains and target domain to detect the informative target samples. Specifically, the leverages a dynamic Domain Adaptation(DA) model that learns how to adapt the model's parameters to fit the union of multi-source domains. This enables an approximate single-source domain modeling by the dynamic model. We then comprehensively measure both domain uncertainty and predictive uncertainty in the target domain to detect informative target samples using evidential deep learning, thereby mitigating uncertainty miscalibration. Furthermore, we introduce a contextual diversity-aware calculator to enhance the diversity of the selected samples. Experiments demonstrate that our solution outperforms existing methods by a considerable margin on three domain adaptation benchmarks.

cs.AI↗

arXiv2023

RECALL: A Benchmark for LLMs Robustness against External Counterfactual Knowledge

LLMs and AI chatbots have improved people's efficiency in various fields. However, the necessary knowledge for answering the question may be beyond the models' knowledge boundaries. To mitigate this issue, many researchers try to introduce external knowledge, such as knowledge graphs and Internet contents, into LLMs for up-to-date information. However, the external information from the Internet may include counterfactual information that will confuse the model and lead to an incorrect response. Thus there is a pressing need for LLMs to possess the ability to distinguish reliable information from external knowledge. Therefore, to evaluate the ability of LLMs to discern the reliability of external knowledge, we create a benchmark from existing knowledge bases. Our benchmark consists of two tasks, Question Answering and Text Generation, and for each task, we provide models with a context containing counterfactual information. Evaluation results show that existing LLMs are susceptible to interference from unreliable external knowledge with counterfactual information, and simple intervention methods make limited contributions to the alleviation of this issue.

cs.CL↗

arXiv2023

Heuristic Algorithms for RIS-assisted Wireless Networks: Exploring Heuristic-aided Machine Learning

Reconfigurable intelligent surfaces (RISs) are a promising technology to enable smart radio environments. However, integrating RISs into wireless networks also leads to substantial complexity for network management. This work investigates heuristic algorithms and applications to optimize RIS-aided wireless networks, including greedy algorithms, meta-heuristic algorithms, and matching theory. Moreover, we combine heuristic algorithms with machine learning (ML), and propose three heuristic-aided ML algorithms, namely heuristic deep reinforcement learning (DRL), heuristic-aided supervised learning, and heuristic hierarchical learning. Finally, a case study shows that heuristic DRL can achieve higher data rates and faster convergence than conventional deep Q-networks (DQN). This work provides a new perspective for optimizing RIS-aided wireless networks by taking advantage of heuristic algorithms and ML.

cs.NI↗

arXiv2023

Rethinking Evaluation Metrics of Open-Vocabulary Segmentaion

In this paper, we highlight a problem of evaluation metrics adopted in the open-vocabulary segmentation. That is, the evaluation process still heavily relies on closed-set metrics on zero-shot or cross-dataset pipelines without considering the similarity between predicted and ground truth categories. To tackle this issue, we first survey eleven similarity measurements between two categorical words using WordNet linguistics statistics, text embedding, and language models by comprehensive quantitative analysis and user study. Built upon those explored measurements, we designed novel evaluation metrics, namely Open mIoU, Open AP, and Open PQ, tailored for three open-vocabulary segmentation tasks. We benchmarked the proposed evaluation metrics on 12 open-vocabulary methods of three segmentation tasks. Even though the relative subjectivity of similarity distance, we demonstrate that our metrics can still well evaluate the open ability of the existing open-vocabulary segmentation methods. We hope that our work can bring with the community new thinking about how to evaluate the open ability of models. The evaluation code is released in github.

cs.CV↗