Bryan Hooi lists you as their PhD student. Claim this profile to confirm it and keep the rest of your record right.
Claim this profileAcademic lineage
View as a treeAdvisors
Works290 from public data
- Co-Occurrence Feature Learning for Skeleton Based Action Recognition Using Regularized Deep LSTM Networks930
This work takes the skeleton as the input at each time slot and introduces a novel regularization scheme to learn the co-occurrence features of skeleton joints, and proposes a new dropout algorithm which simultaneously operates on the gates, cells, and output responses of the LSTM neurons.
- A Sufficient Condition for Convergences of Adam and RMSProp470
An alternative easy-to-check sufficient condition is introduced, which merely depends on the parameters of the base learning rate and combinations of historical second-order moments, to guarantee the global convergence of generic Adam/RMSProp for solving large-scale non-convex stochastic optimization.
- PyTorch distributed468
Evaluations show that, when configured appropriately, the PyTorch distributed data parallel module attains near-linear scalability using 256 GPUs.
- 339
- 328
- 270
- EdgeFed: Optimized Federated Learning Based on Edge Computing197
Inspired by edge computing, this work proposed edge federated learning (EdgeFed), which separates the process of updating the local model that is supposed to be completed independently by mobile devices, to improve the learning efficiency and decrease the global communication frequency.
- 192
- DisPFL: Towards Communication-Efficient Personalized Federated Learning via Decentralized Sparse Training186
This work proposes a novel personalized federated learning framework in a decentralized (peer-to-peer) communication protocol named Dis-PFL, which employs personalized sparse masks to customize sparse local models on the edge to save the communication and computation cost.
- 158
- MiLeNAS: Efficient Neural Architecture Search via Mixed-Level Reformulation147
It is shown that even when using a simple first-order method on the mixed-level formulation, MiLeNAS can achieve a lower validation error for NAS problems, and architectures obtained by the method achieve consistently higher accuracies than those obtained from bilevel optimization.
- 146
- A Comprehensive Survey of Forgetting in Deep Learning Beyond Continual Learning139
This survey argues that forgetting is a double-edged sword and can be beneficial and desirable in certain cases, such as privacy-preserving scenarios, and presents a more nuanced understanding of this phenomenon and highlights its potential advantages.
- Deep Model Fusion: A Survey134
A comprehensive survey to summarize the recent progress of deep model fusion/merging and analyzes the challenges of deep model fusion to illuminate the possible research directions in the future.
- Stochastic Client Selection for Federated Learning With Volatile Clients133
This article investigates the client selection problem under a volatile context, in which the local training of heterogeneous clients is likely to fail due to various kinds of reasons and in different levels of frequency, and proposes E3CS, a stochastic client selection scheme as a solution.
- Learning from models beyond fine-tuning114
This Review provides a comprehensive overview of current methods based on foundation models from the perspective of LFM, and emphasizes that maximizing the use of parametric knowledge in data-scarce scenarios is critical to LFM.
- Make Landscape Flatter in Differentially Private Federated Learning109
A novel DPFL algorithm named DP-FedSAM is proposed, which leverages gradient perturbation to mitigate the negative impact of DP, and empirically confirms that the algorithm achieves state-of-the-art (SOTA) performance compared with existing SOTA baselines in DPFL.
- 105
- Application of the adaptive reuse potential model in Hong Kong: A case study of Lui Seng Chun105
This paper explores the relationships between financial, environmental and social parameters associated with building adaptive reuse by way of a case study of a heritage building in Hong Kong known as Lui Seng Chun.
- On Transforming Reinforcement Learning With Transformers: The Development Trajectory103
This survey collects and dissects recent advances concerning the transformation of RL with transformers (transformer-based RL (TRL), and examines the main applications of TRL in robotic manipulation, text-based games (TBGs), navigation, and autonomous driving.
- Improving the Model Consistency of Decentralized Federated Learning103
Two DFL algorithms named DFedSAM and DFed SAM-MGS are proposed to improve the performance of DFL and can achieve competitive performance compared with CFL methods and outperform existing DFL methods.
- 103
- On Efficient Training of Large-Scale Deep Learning Models102
This survey presents a comprehensive review of the general techniques for training acceleration and provides a detailed analysis and discussion of future works on the development of general acceleration techniques, which could inspire us to re-think and design novel efficient paradigms.
- FedSpeed: Larger Local Interval, Less Communication Round, and Higher Generalization Accuracy90
The proposed FedSpeed merges the vanilla stochastic gradient with a perturbation computed from an extra gradient ascent step in the neighborhood, thereby alleviating the issue of local over-fitting and achieves the state-of-the-art (SOTA) performance on the general FL experimental settings than several baselines.
- Dynamic Regularized Sharpness Aware Minimization in Federated Learning: Approaching Global Consistency and Smooth Landscape88
This work proposes a novel and general algorithm that adopts a dynamic regularizer to guarantee the local optima towards the global objective, which is meanwhile revised by the global Sharpness Aware Minimization (SAM) optimizer to search for the consistent flat minima.
- 88
- 80
- Comparator Networks77
A Deep Comparator Network is proposed that can ingest a pair of sets, and compute a similarity between the pair, and is Inspired by image retrieval, a novel hard sample mining regime is proposed to control the sampling process, such that the DCN is complementary to the standard image classification models.
- Achievable Information Rates for Nonlinear Fiber Communication via End-to-end Autoencoder Learning76
The approach jointly optimizes the input distribution (constellation shaping) and the auxiliary channel distribution to compute AIRs without explicit channel knowledge in an end-to-end fashion.
- ANN-Based Mark-Up Estimation System with Self-Explanatory Capacities73
The research presented in this paper investigates the use of the KT-1 method for automatically extracting rules from a trained neural network for mark-up estimation and indicates the usefulness of the method.
- 72
- Pilot study: Application of artificial intelligence for detecting left atrial enlargement on canine thoracic radiographs71
Although overall accuracy of the accuracy driven convolutional neural network algorithm and veterinary radiologists was identical, concordance between the two approaches was 85.19%.
- 65
- 62
- Initializing Convolutional Filters with Semantic Features for Text Classification62
This paper presents a novel weight initialization method to improve the CNNs for text classification by encoding semantic features into them, which helps the model focus on learning useful features at the beginning of the training.
- 61
- Scalable social sensing of interdependent phenomena59
Evaluation shows that combining expected correlation graphs (of outages) with reported observations of unknown reliability, results in a much more reliable reconstruction of ground truth from the noisy social sensing data, and shows that correlation graphs can help test hypotheses regarding underlying causes, when different hypotheses are associated with different correlation patterns.
- WisdoM: Improving Multimodal Sentiment Analysis by Fusing Contextual World Knowledge58
A plug-in framework named WisdoM, to leverage the contextual world knowledge induced from the large vision-language models (LVLMs) for enhanced MSA, and to reduce the noise in the context, a training-free contextual fusion mechanism is designed.
- A framework for implementing ISO 14000 in construction57
It is suggested that the adoption of ISO 14000 by construction companies may enable them to improve their environmental performance as well as the built environment, which in turn will contribute to sustainable development.
- RDeepSense54
RDeepSense is proposed, the first deep learning model that provides well-calibrated uncertainty estimations for resource-constrained mobile and embedded devices and can reduce around 90% of the energy consumption and preserve at least the same model accuracy compared with other state-of-the-art methods.
- LGViT: Dynamic Early Exiting for Accelerating Vision Transformer53
A novel two-stage training scheme is developed, including end-to-end training and self-distillation with the backbone frozen to generate early exiting ViTs, which facilitates the fusion of global and local information extracted by the two types of heads.
- AlphaGAN: Fully Differentiable Architecture Search for Generative Adversarial Networks53
This work proposes a fully differentiable search framework, dubbed alphaGAN, where the searching process is formalized as solving a bi-level minimax optimization problem, and provides a comprehensive analysis of the behavior of the search process and the properties of searched architectures, which would benefit further research on architectures for generative models.
- DEAL: An Unsupervised Domain Adaptive Framework for Graph-level Classification52
A novel GNN framework named DEAL is presented by incorporating both source graphs and target graphs, which is featured by two modules, i.e., adversarial perturbation and pseudo-label distilling, which improves the performance of graph classification.
- 51
- 51
- Targeted Vaccine: Safety Alignment for Large Language Models Against Harmful Fine-Tuning via Layer-Wise Perturbation50
A Targeted Vaccine (T-Vaccine), a memory-efficient safety alignment method that applies perturbation to only selected layers of the model, which outperforms Vaccine in terms of both defense effectiveness and resource efficiency.
- Using Machine Learning Techniques to Develop Risk Prediction Models for the Risk of Incident Diabetic Retinopathy Among Patients With Type 2 Diabetes Mellitus: A Cohort Study47
The proposed model achieved high performance in predicting the risk of DR among patients with type 2 diabetes mellitus at each time point and established the potential of the XGBoost model to facilitate clinicians in identifying high-risk patients and making type 1 diabetes management-related decisions.
- 47
- Efficient Federated Learning Via Local Adaptive Amended Optimizer With Linear Speedup46
This work incorporates a locally amended technique to the adaptive optimizer, named Federated Local ADaptive Amended optimizer (FedLADA), which estimates the global average offset in the previous communication round and corrects the local offset through a momentum-like term to further improve the empirical training speed and mitigate the heterogeneous over-fitting.
- Dynamic Contrastive Distillation for Image-Text Retrieval45
A novel plug-in dynamic contrastive distillation (DCD) framework to compress the large VLP models for the ITR task and proposes dynamic distillation to dynamically learn samples of different difficulties to balance better the difficulty of knowledge and students' self-learning ability.
- Subspace based Federated Unlearning44
This paper proposes a simple-yet-effective subspace based federated unlearning method, dubbed SFU, that lets the global model perform gradient ascent in the orthogonal space of input gradient spaces formed by other clients to eliminate the target client's contribution without requiring additional storage.
- 44
- Concrete Subspace Learning based Interference Elimination for Multi-task Model Fusion43
This work proposes the CONtinuous relaxation of disCRETE (Concrete) subspace learning method to identify a common low-dimensional subspace and utilize its shared information to track the interference problem without sacrificing much performance.
- 43
- 42
- Enhanced Balanced Min Cut42
In the new method, a new normalized cut model is proposed, in which a set of balance parameters are learned to capture the differences among different clusters, in order to effectively solve the new model without eigendecomposition.
- Weighted AdaGrad with Unified Momentum42
AdaUSM is proposed, which has the main characteristics that it incorporates a unified momentum scheme which covers both the heavy ball momentum and the Nesterov accelerated gradient momentum, and adopts a novel weighted adaptive learning rate that can unify the learning rates of AdaGrad, AccAdaGrad, Adam, and RMSProp.
- Joint Admission Control and Resource Allocation of Virtual Network Embedding via Hierarchical Deep Reinforcement Learning41
A deep Hierarchical Reinforcement Learning approach to learn a joint Admission Control and Resource Allocation policy for VNE, named HRL-ACRA, and develops a deep feature-aware graph neural network to capture the features of VNR and physical network and exploit a sequence-to-sequence model to generate embedding actions iteratively.
- 41
- Quantized Adam with Error Feedback41
A distributed variant of an adaptive stochastic gradient method for training deep neural networks in the parameter-server model is presented and it is shown that the distributed adaptive gradient method with gradient quantization and error feedback converges to the first-order stationary point.
- 40
- 40
- Efficient-Adam: Communication-Efficient Distributed Adam39
This work presents a novel communication-efficient distributed Adam in the parameter-server model for stochastic nonconvex optimization, dubbed Efficient Adam, and incorporates a two-way quantization scheme into Efficient-Adam to reduce the communication cost between the workers and server.
- Pyro: a spatial-temporal big-data storage system36
Pyro is presented, a spatial-temporal big data storage system tailored for high resolution geometry queries and dynamic hotspots, which reduces the response time by 60× on 1km×1km rectangle geometries compared to the state-of-the-art solutions.
- 34
- Differentiable Neural Architecture Search for Extremely Lightweight Image Super-Resolution34
This work proposes a novel differentiable Neural Architecture Search (NAS) approach on both the cell-level and network-level to search for lightweight SISR models, focusing on the combinations of lightweight operations and aiming to build a more lightweight and accurate SR structure.
- Can Linguistic Knowledge Improve Multimodal Alignment in Vision-Language Pretraining?32
This study aims to shed light on the influence of comprehensive linguistic knowledge encompassing semantic expression and syntactic structure on multimodal alignment in vision-language pretraining by introducing SNARE, a large-scale multimodal alignment probing benchmark designed specifically for the detection of vital linguistic components.
- 31
- Proximity-Informed Calibration for Deep Neural Networks30
ProCal is proposed, a plug-and-play algorithm with a theoretical guarantee to adjust sample confidence based on proximity bias that is effective in addressing proximity bias and improving calibration on balanced, long-tail, and distribution-shift settings under four metrics over various model architectures.
- Curriculum-Based Asymmetric Multi-Task Reinforcement Learning30
CAMRL, the first curriculum-based asymmetric multi-task learning (AMTL) algorithm for dealing with multiple reinforcement learning (RL) tasks altogether, is introduced and the improvements of CAMRL over the corresponding single-task RL algorithm and state-of-the-art MTRL algorithms are shown.
- Efficient and Effective Weight-Ensembling Mixture of Experts for Multi-Task Model Merging29
An efficient-and-effective WEMoE (E-WEMoE) method, whose core mechanism involves eliminating non-essential elements in the critical modules of WEMoE and implementing shared routing across multiple MoE modules, thereby significantly reducing both the trainable parameters, the overall parameter count, and computational overhead of the merged model by WEMoE.
- Informative pairs mining based adaptive metric learning for adversarial domain adaptation28
This paper proposes an informative pairs mining based adaptive metric learning (IPM-AML), where a novel two-triplet-sampling strategy is advanced to select informative positive pairs from the same classes and informative negative pairs from different classes, and a metric loss imposed with special weights is further utilized to adaptively pay more attention to those more informative pairs which can adaptively improve discrimination.
- Understanding How Consistency Works in Federated Learning via Stage-wise Relaxed Initialization27
An efficient FL algorithm is designed, which allows employing the personalized relaxed initialization state at the beginning of each local training stage, which initializes the local state by moving away from the current global state towards the reverse direction of the latest local state.
- 27
- 27
- SGDA: Towards 3-D Universal Pulmonary Nodule Detection via Slice Grouped Domain Attention26
A slice grouped domain attention (SGDA) module is proposed to enhance the generalization capability of the pulmonary nodule detection networks and enables substantially better multi- domain pulmonary nodules detection performance compared with the state-of-the-art multi-domain learning methods.
- 26
- Efficient Federated Learning With Enhanced Privacy via Lottery Ticket Pruning in Edge Computing25
The proposed Fed-LTP, an efficient and privacy-enhanced FL framework with Lottery Ticket Hypothesis (LTH) and zero-concentrated DP (zCDP), generates a pruned global model and conducts sparse-to-sparse training from scratch with zCDP on the client side.
- 25
- Discrete Trust-aware Matrix Factorization for Fast Recommendation24
This paper proposes a discrete trust-aware matrix factorization (DTMF) model to take dual advantages of both social relations and discrete technique for fast recommendation, and maps the latent representation of users and items into a joint hamming space by recovering the rating and trust interactions between users and Items.
- Multi-Scenario and Multi-Task Aware Feature Interaction for Recommendation System23
A Multi-scenario and Multi-task aware Feature Interaction model, dubbed MMFI, is proposed to explicitly model feature interactions and learn the importance of feature interaction pairs in different scenarios and tasks, and designs a scenario-task aware attention layer for learning the importance of feature interactions.
- PrimKD: Primary Modality Guided Multimodal Fusion for RGB-D Semantic Segmentation23
This work proposes PrimKD, a knowledge distillation based approach that focuses on guided multimodal fusion, with an emphasis on leveraging the primary RGB modality, and utilizes a model trained exclusively on the RGB modality as the teacher, guiding the learning process of a student model that fuses both RGB and depth modalities.
- Visual Prompt Based Personalized Federated Learning23
A novel PFL framework for image classification tasks, dubbed pFedPT, is proposed that leverages personalized visual prompts to implicitly represent local data distribution information of clients and provides that information to the aggregation model to help with classification tasks.
- 22
- Vulnerability-Aware Alignment: Mitigating Uneven Forgetting in Harmful Fine-Tuning22
This work reveals that certain subsets of alignment data are consistently more prone to forgetting during HFT across different fine-tuning tasks, and proposes Vulnerability-Aware Alignment (VAA), which estimates data vulnerability, partitions data into vulnerable groups and encourages balanced learning using a group distributionally robust optimization (Group DRO) framework.
- Fusion of Global and Local Knowledge for Personalized Federated Learning22
This paper employs proper regularization to extract a low-rank global knowledge representation (GKR) and proposes a two-stage proximal-based algorithm named FedSLR, which reduces the number of parameters, and lowers the down-link communication complexity, which are all desirable for federated learning algorithms.
- On the Convergence of AdaGrad with Momentum for Training Deep Neural Networks22
Two new adaptive stochastic gradient methods called AdaHB and AdaNAG are proposed which integrate coordinate-wise AdaGrad with heavy ball momentum and Nesterov accelerated gradient momentum, respectively and are jointly characterized by leveraging a newly developed unified formulation of these two momentum mechanisms.
- SPORT: A Subgraph Perspective on Graph Classification with Label Noise21
This work investigates the problem of graph classification with label noise, which is demanding because of the complex graph representation learning issue and serious memorization of noisy samples, and presents a novel approach called Subgraph Set Network with Sample Selection and Consistency Learning (SPORT) for this problem.
- UniFaceGAN: A Unified Framework for Temporally Consistent Facial Video Editing21
A novel 3D temporal loss constraint is introduced based on the barycentric coordinate interpolation and a region-aware conditional normalization layer is proposed to replace the traditional AdaIN or SPADE to synthesize more context-harmonious results.
- 21
- 21
- 20
- 19
- 18
- Efficient Federated Prompt Tuning for Black-box Large Pre-trained Models18
Federated Black-Box Prompt Tuning proficiently sidesteps memory challenges tied to PTM storage and fine-tuning on local machines, tapping into comprehensive, high-quality, yet private training datasets.
- Machine Learning-Based Production Prediction Model and Its Application in Duvernay Formation18
Sensitivity analysis based on GPR model indicates that condensate-gas ratio, fluid volume, and total organic carbon content are the most important features to cumulative oil production within 1st 6 producing months.
- 18
- Mitigating the Backdoor Effect for Multi-Task Model Merging via Safety-Aware Subspace17
This paper proposes a novel Defense-Aware Merging (DAM) approach that simultaneously mitigates task interference and backdoor vulnerabilities, and achieves a more favorable balance between performance and security.
- 17
- A lightweight cryptographic algorithm for the transmission of images from road environments in self-driving17
This work proposes a security strategy in the edge computing that combines the Feistel architecture and short comparable encryption based on sliding window (SCESW) and guarantees its security while significantly reducing the computational overhead.
- Instructed Diffuser With Temporal Condition Guidance for Offline Reinforcement Learning16
An effective temporally-conditional diffusion model coined Temporally-Composable Diffuser (TCD) is proposed, which extracts temporal information from interaction sequences and explicitly guides generation with temporal conditions.
- Toward the Flatter Landscape and Better Generalization in Federated Learning Under Client-Level Differential Privacy16
A novel DPFL algorithm named DP-FedSAM is proposed, which leverages gradient perturbation to mitigate the negative impact of DP, and gives rigorous privacy guarantees with Rényi DP, the sensitivity analysis of local updates, and generalization analysis.
- 16
- 16
- Don’t Be So Dense: Sparse-to-Sparse GAN Training Without Sacrificing Performance16
Sparsified GANs, trained from scratch in one single run, are able to outperform the ones learned by expensive iterative pruning and re-training, and can achieve even better performance than the dense BigGAN.
- 16
- A Decomposition Algorithm for the Sparse Generalized Eigenvalue Problem16
A new effective decomposition method that uses random or/and swapping strategies to find a working set and perform global combinatorial search over the small subset of variables to solve the sparse generalized eigenvalue problem.
- On the Convergence of Weighted AdaGrad with Momentum for Training Deep Neural Networks16
Two new adaptive stochastic gradient methods are proposed called AdaHB and AdaNAG which integrate a novel weighted coordinate-wise AdaGrad with heavy ball momentum and Nesterov accelerated gradient momentum, respectively.
- Low-Precision Training of Large Language Models: Methods, Challenges, and Opportunities15
This survey provides a comprehensive review of existing low-precision training methods and categorizes them into three primary groups based on their underlying numerical formats, which is a key factor influencing hardware compatibility, computational efficiency, and ease of reference for readers.
- Enhance Local Consistency in Federated Learning: A Multi-Step Inertial Momentum Approach15
A novel federated learning algorithm, named FedMIM, is proposed, which adopts the multi-step inertial momentum on the edge devices and enhances the local consistency for free during the training to improve the robustness of the heterogeneity.
- 15
- 14
- 14
- 14
- 13
- 13
- 13
- Building Accurate Translation-Tailored LLMs with Language Aware Instruction Tuning13
A two-stage fine-tuning algorithm to improve the instruction-following ability (especially the translation direction) of LLMs, which could effectively reduce the off-target translation ratio and improve translation quality.
- Towards More Suitable Personalization in Federated Learning via Decentralized Partial Model Training13
This paper proposes a personalized FL framework with a decentralized partial model training called DFedAlt, which personalizes the "right" components in the modern deep models by alternately updating the shared and personal parameters to train partially personalized models in a peer-to-peer manner and provides convergence analysis of both algorithms in the general non-convex setting.
- Architecture, Dataset and Model-Scale Agnostic Data-free Meta-Learning13
A unified framework, dubbed PURER, is proposed, which contains a unified framework for meta training with ECI as an adversarial form in an end-to-end manner and a simple plug-and-play supplement—ICFIL—only used during meta testing to narrow the gap between meta training and meta testing task distribution.
- 13
- 13
- A Block Decomposition Algorithm for Sparse Optimization13
A new block decomposition algorithm that combines the effectiveness of combinatorial search methods and the efficiency of coordinate descent methods, and generally outperforms the well-known greedy pursuit method is considered.
- Stability and Generalization of the Decentralized Stochastic Gradient Descent Ascent Algorithm12
This paper investigates the primal-dual generalization bound of the decentralized stochastic gradient descent ascent (D-SGDA) algorithm using the approach of algorithmic stability under both convex-concave and nonconvex-nonconCave settings and demonstrates that the decentralized structure does not destroy the stability and generalization of D-SG DA, implying that it can generalize as well as the vanilla SGDA in certain situations.
- Prediction of Online Psychological Help-Seeking Behavior During the COVID-19 Pandemic: An Interpretable Machine Learning Method12
The interpretable model showed that the increase in the OPHS behaviors was impacted by the cumulative confirmed cases and cumulative deaths, while it was not sensitive in the new confirmed cases or new deaths, which demonstrated the power of utilizing the OMHS platforms as an always-on data source to obtain a high-resolution timeline and real-time prediction of the psychological response of the online public.
- 12
- 12
- Adaptive Activation Network and Functional Regularization for Efficient and Flexible Deep Multi-Task Learning11
A novel deep learning model called Task Adaptive Activation Network (TAAN) is proposed that can automatically learn the optimal network architecture for MTL and two functional regularization methods are proposed that improve the MTL performance of TAAN.
- Graphs in Biomedical Image Analysis, Computational Anatomy and Imaging Genetics11
This book constitutes the refereed joint proceedings of the First International Workshop on Graphs in Biomedical Image Analysis, GRAIL 2017, and the 6th International workshop on Mathematical Foundations of Computational Anatomy, MFCA 2017, held in conjunction with the 20th International Conference on Medical Imaging and Computer-Assisted Intervention, MICCAI, in Quebec City, QC, Canada, in September 2017.
- LOST: Low-rank and Sparse Pre-training for Large Language Models10
A novel method that ingeniously integrates low-rank and sparse structures to enable effective training of LLMs from scratch under strict efficiency constraints, and achieves competitive or superior performance compared to full-rank models, while significantly reducing both memory and compute overhead.
- Graph decision transformer for offline reinforcement learning10
This framework unifies multiple components in offline RL, such as estimating the behavior policy and predictive dynamics modeling, into a single sequence model, resulting in superior performance.
- Simultaneous Computation and Memory Efficient Zeroth-Order Optimizer for Fine-Tuning Large Language Models10
A novel layer-wise sparse computation and memory efficient ZO optimizer, named LeZO, which accelerates training without compromising the performance of ZO optimization and incorporates layer-wise parameter sparsity in the process of simultaneous perturbation stochastic approximation and ZO stochastic gradient descent.
- 10
- Super-Resolution Model Quantized in Multi-Precision10
In quantization, it is found that some stages of the two super-resolution model networks, SRGAN and ESRGAN, showed sensitivity toquantization, which greatly reduced the performance, so higher-bits integer quantization is used for the sensitive stage, and the accuracy approaching the original model was achieved.
- Incorporating URL embedding into ensemble clustering to detect web anomalies10
A weighted deep learning enabled subspace spectral ensemble clustering approach for web anomaly detection called WDL-SSEC achieves better performance than existing approaches, which demonstrates that the model has the ability to cluster anomalies into appropriate types.
- Omniforce: on human-centered, large model empowered and cloud-edge collaborative AutoML system9
OmniForce is introduced, a human-centered AutoML system that yields both human-assisted ML and ML-assisted human techniques, putting an AutoML system into practice and building AI in open-environment scenarios, such as using large models in industrial supply chains.
- DFedADMM: Dual Constraints Controlled Model Inconsistency for Decentralized Federated Learning9
Novel DFL algorithms are proposed that exhibit superior performance in terms of both generalization and convergence speed compared to existing state-of-the-art (SOTA) optimizers in DFL.
- 9
- Offline Quantum Reinforcement Learning in a Conservative Manner9
The first offline quantum RL (offline QRL) algorithm named CQ2L (Conservative Quantum Q-learning), which learns from offline samples and does not require any interaction with the environment, is developed.
- Programmable Synthetic Protein Circuits for the Identification and Suppression of Hepatocellular Carcinoma9
This study designed and verified the basic control layer of protein-protein interaction (PPI) sensing and constructed synthetic-promoter-free gene circuits possessing two control layers to improve the specific recognition of tumor cells.
- 9
- Virus propagation power of the dynamic network9
This study offers a feasible approach for quantifying the risk of virus infection in the network community which is valuable for designing and optimizing the virus defense systems.
- 9
- Developing an optimal bidding strategy in two-envelope fee bidding9
This paper offers consultants an approach to better identify their optimum technical score+fee combination, which is based on the consultant's original technical proposal and fee, and the absolute highest technical score and corresponding fee.
- 8
- DFedGFM: Pursuing global consistency for Decentralized Federated Learning via global flatness and global momentum8
This paper first rethink the local flatness and local momentum acceleration techniques in the two most popular DFL optimizers and proposes a novel DFL algorithm, called DFedGFM, to simultaneously explore the global flatness and global momentum to achieve global consistency.
- 8
- 8
- From Random to Supervised: A Novel Dropout Mechanism Integrated with Global Information8
GI-Dropout is presented, a novel dropout method integrating with global information to improve neural networks for text classification that aims to use explicit instructions based on global information of the dataset to guide the training process.
- 8
- An Internet‐based electronic product catalogue of construction materials8
This paper attempts to review problems of existing electronic product catalogues on construction material websites and presents an electronic product catalogue designed to facilitate easy searching, comparing and evaluating of construction materials.
- CoFormer: Collaborating With Heterogeneous Edge Devices for Scalable Transformer Inference7
CoFormer enables the efficient inference of GPT2-XL with 1.6 billion parameters on edge devices, reducing memory requirements by 76.3%.
- ADEM-VL: Adaptive and Embedded Fusion for Efficient Vision-Language Tuning7
ADEM-VL is proposed, an efficient vision-language method that tunes VL models based on pretrained large language models by adopting a parameter-free cross-attention mechanism for similarity measurements in multimodal fusion that outperforms existing approaches with reduced training and inference latency.
- Dynamic Analysis and Adaptive Discriminator for Fake News Detection7
The Monte Carlo Tree Search algorithm is introduced to leverage the self-reflective capabilities of large language models (LLMs) for prompt optimization, providing richer, domain-specific details and guidance to the LLMs, while enabling more flexible integration of LLM comment on news content.
- Sequential Federated Learning in Hierarchical Architecture on Non-IID Datasets7
For the first time, SFL is introduced into HFL and a novel algorithm adapted to this combined framework, called Fed-CHS is proposed, which provides evidence of the superiority of the superiority of the proposed Fed-CHS on both communication overhead saving and test accuracy over baseline methods.
- Cross-Domain Diffusion With Progressive Alignment for Efficient Adaptive Retrieval7
This approach revisits unsupervised efficient domain adaptive retrieval from a graph diffusion perspective, simulating cross-domain adaptation dynamics to achieve a stable target domain adaptation process and enables effective domain adaptive hash learning.
- Systematic Investigation of Sparse Perturbed Sharpness-Aware Minimization Optimizer7
This paper proposes Sparse SAM (SSAM), an efficient and effective training scheme that achieves sparse perturbation by a binary mask and theoretically proves that SSAM can converge at the same rate as SAM.
- Continual Learning From a Stream of APIs7
A novel data-free cooperative continual distillation learning framework that distills knowledge from a stream of APIs into a CL model by generating pseudo data, just by querying APIs, and proposes a new regularization term based on network similarity to prevent catastrophic forgetting of previous APIs.
- Federated Learning With Only Positive Labels by Exploring Label Correlations7
This article proposes a novel and generic method termed federated averaging (FedAvg) by exploring label correlations (FedALCs), which estimates the label correlations in the class embedding learning for different label pairs and utilizes it to improve the model training.
- Local AdaGrad-type algorithm for stochastic convex-concave optimization7
A communication-efficient distributed stochastic extragradient algorithm, LocalAdaSEG, with an adaptive learning rate suitable for solving convex-concave minimax problems in the Parameter-Server model, used to solve a stochastic bilinear game, and train a generative adversarial network.
- 7
- 6
- Task-Distributionally Robust Data-Free Meta-Learning6
This paper proposes a trustworthy DFML framework comprising three components: synthetic task reconstruction, meta-learning with task memory interpolation, and automatic model selection, which seamlessly incorporates an automatic model selection mechanism to automatically filter out untrustworthy models during the meta-learning process.
- 6
- Quantum Imitation Learning6
This work is the first to propose the concept of QIL and conduct pilot studies, which paves the way for the quantum era and demonstrates that both Q-BC and Q-GAIL can achieve comparable performance compared to classical counterparts, with the potential of quantum speedup.
- 6
- Improved uniqueness conditions of solution for multilinear pagerank and its application6
Several improved uniqueness conditions of solution for multilinear PageRank with order-3 irreducible stochastic tensors are proposed by considering upper and lower bounds of solutions and are more effective than the existing ones.
- Adaptive Proximal Average Approximation for Composite Convex Minimization6
Preliminary experimental results on overlapping group Lasso and graph-guided fused Lasso problems indicate that the proposed method is almost five times faster than the state-of-the-art PA-APG method and therefore suitable for higher-precision required optimization.
- 6
- GSOS: Gauss-Seidel Operator Splitting Algorithm for Multi-Term Nonsmooth Convex Composite Optimization6
The proposed GSOS algorithm inherits the advantage of the Gauss-Seidel technique to accelerate the optimization procedure, and leverages the operator splitting technique to reduce the computational complexity.
- 6
- 6
- Comparing two processing routines for GPS traces: Lessons learnt6
This paper describes what may be one of the first side-by-side tests of two alternative software products for processing GPS traces into trips, and discusses some lessons learnt from the comparisons.
- 6
- 5
- 5
- 5
- DGL-GAN: discriminator-guided GAN compression5
This work proposes a novel yet simple Discriminator-Guided Learning approach for compressing vanilla GAN, dubbed DGL-GAN, and applies it to compress the two most representative large-scale vanilla GANs, i.e., StyleGAN2 and BigGAN.
- Trust, but Verify: Using Self-supervised Probing to Improve Trustworthiness5
This paper introduces a new approach of self-supervised probing, which enables to check and mitigate the overconfidence issue for a trained model, thereby improving its trustworthiness and provides a simple yet effective framework, which can be flexibly applied to existing trustworthiness-related methods in a plug-and-play manner.
- Enhancing Top-N Item Recommendations by Peer Collaboration5
This paper proposes a top-N item recommendation framework called PCRec that leverages collaborative training of two recommender models of the same network structure, termed peer collaboration, and introduces two criteria to identify the importance of parameters of a given recommender model.
- RT Engine: An Efficient Hardware Architecture for Ray Tracing5
A novel architecture, a RT engine (Ray Tracing engine), that accelerates ray tracing is presented, which achieves a performance per area which is 2.4 × greater than the best reported results for ray tracing on dedicated hardware.
- 5
- MAP Inference Via $$\ell _2$$-Sphere Linear Program Reformulation5
A perturbed alternating direction method of multipliers (ADMM) algorithm is proposed to optimize the LS–LP problem, by adding a sufficiently small perturbation of the original MAP inference problem, and it is proved that the perturbed ADMM algorithm globally converges to the ϵ\documentclass[12pt]{minimal} \usepackage{amsmath} \use package{wasysym} ℓ2.
- 5
- 5
- 5
- Toward Understanding the Generalizability of Delayed Stochastic Gradient Descent4
This paper investigates sharper generalization error bound for SGD with asynchronous delay with upper bounds on the generalization error of <inline-formula><tex-math notation="LaTeX">$\widetilde{\mathcal {O}}(\frac{1}{n})$</tex-math><alternatives><mml:math><mml:mrow><mml:mrow><mml:mover accent="true"><mml:mi mathvariant="script
- 4
- 4
- Unlocking Tuning-Free Few-Shot Adaptability in Visual Foundation Models by Recycling Pre-Tuned LoRAs4
This work explores the potential of reusing diverse pre-tuned LoRAs without accessing their original training data, to achieve tuning-free few-shot adaptation in VFMs and distills a meta-LoRA from diverse pre-tuned LoRAs with a meta-learning objective, using surrogate data generated inversely from pre-tuned LoRAs themselves.
- 4
- 4
- 4
- 4
- 4
- Fuzzy logic control ASIC chip4
A fuzzy logic control VLSI chip, F100, for industry process real-time control has been designed and fabricated with 0.8μm CMOS technology and the fuzzy control inference method is presented.
- VLSI Functional Testing Using Critical Path Traces at a Hardware Description Language Level4
A new approach for testing VLSI circuits is presented through backward critical path tracing, a test and all faults detectable by the test are generated simultaneously, therefore, the expensive fault simulation is completely eliminated.
- 3
- Toward Understanding Generalization and Stability Gaps Between Centralized and Decentralized Federated Learning3
It is proved that on the general smooth non-convex objectives, centralized FL (CFL) always generalizes better than decentralized FL (DFL); and there is a necessary requirement for the topology in DFL to avoid performance collapse as the training scale increases.
- Data-Adaptive Weight-Ensembling for Multi-task Model Fusion3
This work proposes a data-adaptive weight-ensembling approach that generates model weights in time for unified model fusion, demonstrating superior performance and satisfactory zero-shot transferability on vision Transformers and Flan-T5 models.
- 3
- Generalized Embedding Machines for Recommender Systems3
This work proposes an alternative approach to model high-order interaction signals at the embedding level, namely generalized embedding machine (GEM), and utilizes graph convolution networks (GCN) to generate high-order embeddings.
- FedLALR: Client-Specific Adaptive Learning Rates Achieve Linear Speedup for Non-IID Data3
This paper proposes a heterogeneous local variant of AMSGrad, named FedLALR, in which each client adjusts its learning rate based on local historical gradient squares and synchronized learning rates, which enables promising scalability in federated optimization.
- Off-policy Imitation Learning from Visual Inputs3
OPIfVI is proposed, which is composed of an off-policy learning manner, data augmentation, and encoder techniques, to tackle the mentioned challenges, respectively, and can achieve expert-level performance and outperform existing baselines via extensive experiments using DeepMind Control Suite.
- Personalized Recommendation Algorithm Based on User Preference and User Profile3
The experimental results show that the personalized recommendation algorithm based on user profile and user preference model can effectively improve the performance of the recommendation system.
- 3
- 3
- 3
- 2
- 2
- Mask-Enhanced Autoregressive Prediction: Pay Less Attention to Learn More2
The analysis indicates that MEAP's effectiveness arises from its ability to promote more distinguishable attention scores by concentrating on a reduced set of non-masked tokens, which improves the model's focus on task-relevant signals while mitigating the influence of peripheral context.
- 2
- Exploring Semantic Topics in Dementia Caregiver Tweets2
This study performs a novel topic model leveraging weighting strategies to analyze dementia‐caregiver‐related twitter data and shows how topic modeling has become an efficient NLP tool for analyzing semantic themes within texts.
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 1
- 1
- 1
- Neural-aware Decoupling Fusion based Personalized Federated Learning for Intelligent Sensing1
A new explainable neural-aware decoupling fusion based PFL framework, p-FedADF, to address the preceding challenges and provides a mask encoding scheme to reduce the communication overhead of transmitting the sub-network sets between the server and clients.
- 1
- 1
- 1
- 1
- 1
- An Improved Training Sample Set Construction Method1
By setting a reasonable threshold value of Euclidean distance similarity, the proposed method can ensure that the constructed sample set have structural anisotropy and diversity and the quality of image reconstruction is improved.
- Control Divergence Optimization through Partial Warp Regrouping in GPGPUs1
This paper proposes a lightweight warp regrouping method-Partial Warp Regrouping (PWR) that avoids most of the unnecessary warp regroupsing by setting thresholds and in this method, it also can reduce the complexity of hardware design.
- 1
- 1
- 1
- 1
- 1
- Learning-to-Share Based on Finding Groups for Large Scale Image Classification1
A method to learn a group-based sharing model such that the visually similar classes are assigned to a discriminative group, which enables the class draw support from other classes in the same group, thus the poor discrimination ability with limited available samples can be relieved.
- 1
- VFSim: Concurrent Fault Simulation at Register Transfer Level1
A Verilog Register transfer level Model (VRM) for integrated circuits is proposed and an RTL concurrent fault simulation approach is presented, which shows that the RTL fault simulator is efficient for VLSI circuits.
- –
- COMPETITION PRACTICE AND RESEARCH OF PROMOTING TEACHING BY COMPETITION: TAKE THE COURSE OF INFORMATION SYSTEM AND SECURITY COUNTERMEASURES TECHNOLOGY FOR AN EXAMPLE–
The research results demonstrate that the "competition-driven teaching" model can achieve a virtuous cycle of "teaching-competition-research-industry," providing theoretical reference for teaching reforms in network space security courses.
- –
- –
- ScaleNet: Scaling up Pretrained Neural Networks With Incremental Parameters–
ScaleNet facilitates rapid model expansion with negligible increases in parameters, building on existing pretrained models, and shows significant potential for application in downstream vision areas, as evidenced by the validation in object detection task.
- –
- –
- –
- Dynamic PDGAN: discriminator-boosted knowledge distillation for StyleGANs–
This work proposes to exploit the teacher discriminator to additionally supervise the student generator of StyleGANs so as to leverage the knowledge in the teacher discriminator, and can outperform original StyleGANs on several large-scale datasets.
- –
- Design and implementation of an advanced grid security event analysis system–
The weaknesses and deficiencies in the current management of grid security events are analyzed and a design for an intelligent analysis system for these events is proposed and coupling with external influencing factors is realized, facilitating the analysis of characteristics of various types of grid security events.
- Heterogeneous Federated Learning with Splited Language Model–
This paper harnesses Pre-trained Image Transformers (PITs) as the initial model, coined FES-PIT, to accelerate the training process and improve model robustness, and proposes FES-PTZO to hinder the gradient inversion attack.
- Master–Slave Deep Architecture for Top-K Multiarmed Bandits With Nonlinear Bandit Feedback and Diversity Constraints–
This work proposes a novel master–slave architecture to solve the top- $K$ combinatorial multiarmed bandits problem with nonlinear bandit feedback and diversity constraints, which is the first combinatorial bandits setting considering diversity constraints under bandit feedback.
- –
- –
- –
- –
- –
- The Algorithm of Direct Current Smooth Filtering Based on the Minimum Resource Consumption in FPGA–
The algorithm of direct current smooth filtering was proposed in this paper to satisfy the requirement of minimum resource consumption in FPGA and to meet the goal of calculation period in minimum length.
- –
- –
- –
- –
- –
- –
- –
- [Establishment of early warning control charts based on the syndromic surveillance data of outpatient diarrhea in Beijing].–
The enteric clinic requires to further collect baseline data to evaluate and continuously adjust the established control charts for the best early warning model in accordance with the enteringic clinic.
- –
- Quantum GA-PLS for feature selection method and its application–
QGA-PLS has better performance in feature selection, execution time and classification accuracy, which proves the efficiency of proposed method.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- NEEDED PARALLEL-OUTERMOST REDUCTION STRATEGY–
By minimizing needed sets, this strategy is complete for the regular term rewriting systems, is efficient like call by need strategy, and can satisfy all demands of applicability, efficiency andcomputability.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.