Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks544 from public data
- CoolStreaming/DONet: A Data-Driven Overlay Network for Efficient Live Media Streaming635
This paper presents DONet, a Data-driven Overlay Network for live media streaming, and presents an efficient memberand partnership management algorithm, together with an intelligent scheduling algorithm that achieves real-time and continuous distribution of streaming contents.
- OpenOOD: Benchmarking Generalized Out-of-Distribution Detection416
A unified, well-structured codebase called OpenOOD is built, which implements over 30 methods developed in relevant fields and provides a comprehensive benchmark under the recently proposed generalized OOD detection framework.
- Google USM: Scaling Automatic Speech Recognition Beyond 100 Languages412
The Universal Speech Model is introduced, a single large model that performs automatic speech recognition (ASR) across 100+ languages by pre-training the encoder of the model on a large unlabeled multilingual dataset, and fine-tuning on a smaller labeled dataset.
- 399
- Efficient task-specific data valuation for nearest neighbor algorithms317
An Monte Carlo approximation algorithm is proposed, which is up to three orders of magnitude faster than the baseline approximation algorithm and can accelerate the value calculation process even further.
- MIMIC-IT: Multi-Modal In-Context Instruction Tuning316
MultI-Modal In-Context Instruction Tuning (MIMIC-IT), a dataset comprising 2.8 million multimodal instruction-response pairs, with 2.2 million unique instructions derived from images and videos, is presented and a large VLM named Otter is trained.
- 275
- Federated Learning with Label Distribution Skew via Logits Calibration271
This paper investigates the label distribution skew in FL and proposes FedLC, which calibrates the logits before softmax cross-entropy according to the probability of occurrence of each class, and applies a fine-grained calibrated cross-entropy loss to local update by adding a pairwise label margin.
- Video-MMMU: Evaluating Knowledge Acquisition from Multi-Discipline Professional Videos241
Video-MMMU is introduced, a multi-modal, multi-disciplinary benchmark designed to assess LMMs' ability to acquire and utilize knowledge from videos, and proposes a proposed knowledge gain metric, {\Delta}knowledge, that quantifies improvement in performance after video viewing.
- Evaluating ChatGPT's Information Extraction Capabilities: An Assessment of Performance, Explainability, Calibration, and Faithfulness221
ChatGPT's performance in Standard-IE setting is poor, but it surprisingly exhibits excellent performance in the OpenIE setting, as evidenced by human evaluation, and indicates that ChatGPT provides high-quality and trustworthy explanations for its decisions.
- HALC: Object Hallucination Reduction via Adaptive Focal-Contrast Decoding220
HALC, a novel decoding algorithm designed to mitigate OH in LVLMs, is introduced, which leverages distinct fine-grained optimal visual information in vision-language tasks and operates on both local and global contexts simultaneously.
- Cross-Modal Progressive Comprehension for Referring Segmentation202
This paper proposes a cross-modal progressive comprehension (CMPC) scheme to effectively mimic human behaviors and implements it as a CMPC-I (Image) module and aCMPC-V (Video) module to improve referring image and video segmentation models.
- UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning201
UI-TARS-2 is presented, a native GUI-centered agent model that addresses challenges through a systematic training methodology: a data flywheel for scalable data generation, a stabilized multi-turn RL framework, a hybrid GUI environment that integrates file systems and terminals, and a unified sandbox platform for large-scale rollouts.
- PixMix: Dreamlike Pictures Comprehensively Improve Safety Measures200
A new data augmentation strategy utilizing the natural structural complexity of pictures such as fractals is designed, which outperforms numerous baselines, is near Pareto-optimal, and roundly improves safety measures.
- EIA: Environmental Injection Attack on Generalist Web Agents for Privacy Leakage182
This work presents a realistic threat model for attacks on the website, and proposes a novel attack method, termed Environmental Injection Attack (EIA), which injects malicious content designed to adapt well to environments where the agents operate and instantiates EIA specifically for privacy scenarios in web environments.
- Adversary Guided Asymmetric Hashing for Cross-Modal Retrieval167
This paper proposes a novel deep cross-modal hashing method, called Adversary Guided Asymmetric Hashing (AGAH), which employs an adversarial learning guided multi-label attention module to enhance the feature learning part which can learn discriminative feature representations and keep the cross- modal invariability.
- 165
- LINKTELLER: Recovering Private Edges from Graph Neural Networks via Influence Analysis144
This paper proposes a privacy attack LINKTELLER via influence analysis to infer the private edge information held by Alice via designing adversarial queries for Bob, and empirically shows that LINK TELLER is able to recover a significant amount of private edges in different settings, both including inductive and transductive datasets, under different graph densities, significantly outperforming existing baselines.
- On the Tool Manipulation Capability of Open-source Large Language Models139
It is demonstrated that classical methods in LLM literature can adapt as model alignment with programmatic data generation, system prompts and in-context demonstration retrievers to enhance open-source LLMs for tool manipulation, and it is shown that such enhancement typically requires about one developer day to curate data for each tool, rendering a recipe with practical amount of human supervision.
- MixGen: A New Multi-Modal Data Augmentation136
MixGen is presented: a joint data augmentation for vision-language representation learning to further improve data efficiency and generates new image-text pairs with semantic relationships preserved by interpolating images and concatenating text.
- LSTM-Based Analysis of Industrial IoT Equipment135
This paper aims to develop a method of analyzing equipment working condition based on the sensed data and building a prediction model for working status forecasting and designing a deep neural network model to predict equipment running data and improving the prediction accuracy by systematic feature engineering and optimal hyperparameter searching.
- Step-Audio: Unified Understanding and Generation in Intelligent Speech Interaction129
Step-Audio is introduced, the first production-ready open-source solution that achieves state-of-the-art performance in human evaluations, especially in terms of instruction following and on open-source benchmarks like LLaMA Question.
- DiffScene: Diffusion-Based Safety-Critical Scenario Generation for Autonomous Vehicles128
A unified approach to generate high-quality safety-critical scenarios which are both realistic and safety-critical for efficient AV evaluation, and it is demonstrated that training AV algorithms with scenarios generated by DiffScene leads to significantly higher performance in terms of the safety-critical metrics compared to baselines.
- 128
- A comparison of methods for sketch-based 3D shape retrieval127
A more comprehensive comparison of fifteen best sketch-based 3D shape retrieval methods is performed by completing the evaluation of each method on both benchmarks.
- Feature extraction using constrained maximum variance mapping126
Experimental results validate that the proposed CMVM approach is superior to other feature extraction methods, such as linear discriminant analysis (LDA), locality preserving projection (LPP), unsupervised discriminant projection (UDP) and maximum variance projection (MVP).
- Learning Invariant Representations and Risks for Semi-supervised Domain Adaptation123
This paper proposes the first method that aims to simultaneously learn invariant representations and risks under the setting of semi-supervised domain adaptation (Semi-DA), and provides a finite sample bound for both classification and regression problems under Semi-DA.
- 123
- From luna to solar111
Luna is a user-space TCP stack that corresponds the latency of network to the speed of SSD; and Solar, a storage-oriented UDP stack that enables both storage and network hardware accelerations, which are the first step towards a high-speed compute-to-storage network in the "storage disaggregation" architecture.
- 105
- On Image Classification: Correlation v.s. Causality.105
Under the assumption that causal relationships among variables are robust across domains, this paper incorporates causal technique into predictive modeling and proposes a novel Causally Regularized Logistic Regression (CRLR) algorithm by jointly optimize global confounder balancing and weighted logistic regression.
- Deep Semantic-Alignment Hashing for Unsupervised Cross-Modal Retrieval102
A novel Deep Semantic-Alignment Hashing (DSAH) for unsupervised cross-modal retrieval, which sufficiently utilizes the co-occurred image-text pairs and elaborately design a semantic-alignment loss function, which elegantly aligns the similarities between features with those between hash codes.
- 96
- InstructRetro: Instruction Tuning post Retrieval-Augmented Pretraining81
This work continues to pretrain a 43B GPT model on additional 100 billion tokens using the Retro augmentation method by retrieving from 1.2 trillion tokens, and finds that one can ablate the encoder from InstructRetro architecture and directly use its decoder backbone, while achieving comparable results.
- 81
- Deep Supervised Hashing for Multi-Label and Large-Scale Image Retrieval72
This work proposes a novel Deep Multilevel Semantic Similarity Preserving Hashing method (DMSSPH) method to learn compact similarity-preserving binary codes for the huge body of multi-label image data with deep convolutional neural networks.
- Incorporating emergency alarms in reliable wireless process control72
The results demonstrate the first systematic approach to incorporate emergency alarms into wireless process control enables timely emergency handling, while allowing regular feedback control loops to effectively share resources in WSANs during normal operations.
- 71
- Unsupervised Learning of Accurate Siamese Tracking71
A novel unsupervised tracking framework is presented, in which a differentiable region mask is proposed to select features as well as to implicitly penalize tracking errors on intermediate frames, and a mask-guided loss reweighting strategy to assign dynamic weights based on the quality of pseudo labels is proposed.
- Dynamic Key-Value Memory Networks With Rich Features for Knowledge Tracing71
This article proposes a new exercise record representation method, which integrates the features of students’ behavior with those of the learning ability, thereby improving the performance of knowledge tracing.
- 66
- Identification of Colon Cancer Using Multi-Scale Feature Fusion Convolutional Neural Network Based on Shearlet Transform63
A multi-scale feature fusion convolutional neural network (MFF-CNN) based on shearlet transform to identify histopathological image of colon cancer with superior performance opens a new perspectives for real-time, objective and accurate diagnosis of cancer.
- Improved mapping of heavy metals in agricultural soils using machine learning augmented with spatial regionalization indices60
The results demonstrated that random forest (RF) models incorporating SRIs outperformed artificial neural network and support vector regression models, achieving R2 values exceeding 0.96 for eight heavy metals on the test data.
- 58
- 55
- LoRA-FA: Efficient and Effective Low Rank Representation Fine-tuning54
Re-sults show that LoRA-FA can always achieve close fine-tuning accuracy across different tasks compared to full parameter fine-tuning and LoRA, and can reduce the overall memory cost by up to 1.4 × compared to LoRA.
- 53
- Stable Learning via Differentiated Variable Decorrelation52
This paper incorporates the unlabled data from multiple environments into the variable decorrelation framework and proposes a Differentiated Variable Decorrelation (DVD) algorithm based on the clustering of variables that improves the model parameter estimation and the prediction stability over changing distributions.
- Understanding Gradual Domain Adaptation: Improved Analysis, Optimal Path and Beyond51
This work analyzes gradual self-training under more general and relaxed assumptions, and proves a significantly improved generalization bound, which implies the existence of an optimal choice of $T$ that minimizes the generalization error, and also naturally suggests an optimal way to construct the path of intermediate domains so as to minimize the accumulative path length $T\Delta$ between the source and target.
- DeMS: A hybrid scheme of task scheduling and load balancing in computing clusters51
Simulation results demonstrate that the proposed On-Demand scheduling algorithm can significantly reduce the response time of parallel jobs, and QMT and STM are effective for independent-task and dependent-task schedulings, respectively.
- 50
- 49
- 47
- Domain Generalization using Pretrained Models without Fine-tuning46
A novel domain generalization paradigm to better leverage various pretrained models is proposed, named specialized ensemble learning fordomain generalization (SEDGE), which first trains a linear label space adapter upon fixed Pretrained models, which transforms the outputs of the pretrained model to the label space of the target domain.
- Deep Unsupervised Hybrid-similarity Hadamard Hashing46
A simple yet effective unsupervised hashing method, dubbed Deep Unsupervised Hybrid-similarity Hadamard Hashing (DU3H), which tackles issues in an end-to-end deep hashing framework and can maximally satisfy the independence and balance properties of hash codes.
- 45
- 44
- AdvWave: Stealthy Adversarial Jailbreak Attack against Large Audio-Language Models43
A dual-phase optimization method is proposed that addresses gradient shattering, enabling effective end-to-end gradient-based optimization and an adaptive adversarial target search algorithm is developed that dynamically adjusts the adversarial optimization target based on the response patterns of LALMs for specific queries.
- C-RAG: Certified Generation Risks for Retrieval-Augmented Language Models42
C-RAG is proposed, the first framework to certify generation risks for RAG models and it is proved that RAG achieves a lower conformal generation risk than that of a single LLM when the quality of the retrieval model and transformer is non-trivial.
- Controllable Time-Delay Transformer for Real-Time Punctuation Prediction and Disfluency Detection42
A Controllable Time-delay Transformer model that jointly completes the punctuation prediction and disfluency detection tasks in real time and facilitates freezing partial outputs with controllable time delay to fulfill the real-time constraints in partial decoding required by subsequent applications is proposed.
- FSSE: Forward secure searchable encryption with keyed-block chains42
The proposed forward secure SSE scheme is efficient since the keyed-block chain can be implemented by symmetric encryption primitive and the scheme also supports both add and delete operations in one instance.
- An Improved NSGA-II Based on Multi-Task Optimization for Multi-UAV Maritime Search and Rescue under Severe Weather41
It is observed that the proposed INSGA-II-MTO can find a set of shorter total paths and handle the most urgent task in the shortest possible time and is an effective and promising approach to solving multi-UAVs MSR problems to reduce human causalities and property losses.
- 41
- 39
- 39
- 39
- Target-Relevant Knowledge Preservation for Multi-Source Domain Adaptive Object Detection38
TRKP adopts the teacher-student framework, where the multi-head teacher network is built to extract knowledge from labeled source domains and guide the student network to learn detectors in unlabeled target domain, thus benefiting decreasing domain shift when mentoring object detection in the target domain.
- Streaming End-to-End Multilingual Speech Recognition with Joint Language Identification38
The structure of the cascaded-encoder-based recurrent neural network transducer (RNN-T) model is modified by integrating a per-frame language identifier (LID) predictor by leveraging differences in the right-contexts and a streaming implementation of statistics pooling to achieve accurate streaming LID prediction with little extra test-time cost.
- 38
- Perada: Parameter-Efficient Federated Learning Personalization with Generalization Guarantees37
Perada is proposed, a parameter-efficient pFL framework that reduces communication and computational costs and exhibits superior generalization performance under test-time distribution shifts and provides generalization bounds of Perada.
- Multimodal image matching based on Multimodality Robust Line Segment Descriptor36
Experimental results indicate that the proposed Multimodality Robust Line Segment Descriptor (MRLSD) matching method achieves higher precision and repeatability than several state-of-the-art local feature-based multimodality matching methods, and also demonstrates its robustness to multimodal images.
- 36
- 35
- Mobile Data Access35
A framework is proposed in which the mobility-related services are decomposed into several levels so that mobility support is made available to many applications, and would help to lead to a mobility enabled wireless IP infrastructure in which voice service is just an application.
- 34
- 33
- 33
- A secure and lightweight privacy-preserving data aggregation scheme for internet of vehicles33
This paper proposes a secure and lightweight privacy-preserving data aggregation scheme, named SLPDA, for IoV that employs Chinese Remainder Theorem to map multi-dimensional data to one- dimensional data, and uses masking technology to encrypt data, thus roadside units can realize lightweight data aggregation without plaintext.
- Carrier squeezing interferometry: suppressing phase errors from the inaccurate phase shift33
The carrier squeezing interferometry algorithm is proposed to retrieve the phase from interferograms with phase shift errors by rearranged image data, and the error lobe and the phase lobe are separated so the error-free phase can be retrieved by filtering.
- Data Debugging with Shapley Importance over End-to-End Machine Learning Pipelines32
DataScope (ease.ml/datascope), the first system that efficiently computes Shapley values of training examples over an end-to-end ML pipeline, is presented and its applications in data debugging for ML training are illustrated.
- 31
- Approximate Group Fairness for Clustering30
The notion of proportional fairness is refined, and the extent to which relaxed core fairness can be satisfied in metric spaces including line, tree and general metric space is studied, and approximation algorithms accordingly are designed.
- 30
- Coupled Metric Learning for Face Recognition with Degraded Images30
This work formulate the coupled metric learning as an optimization problem and solve it efficiently with a closed-form solution, which can be generally applied to face recognition problems with various degrade images.
- 29
- Gradual Domain Adaptation: Theory and Algorithms29
This work theoretically analyze gradual self-training, a popular GDA algorithm, and provides a significantly improved generalization bound compared with Kumar et al. (2020), and proposes GOAT, an algorithmic framework that can generate intermediate domains in a data-dependent way.
- Identification of soil parent materials in naturally high background areas based on machine learning29
This study attempts to determine the key soil geochemical parameters related to the spatial patterns of lithology and the main factors influencing the geochemical behavior of soil Cd, and ultimately uses them and machine-learning methods to identify CA and BA.
- Online Low-Rank Representation Learning for Joint Multi-Subspace Recovery and Clustering29
Experimental results on typical tasks of subspace recovery and subspace clustering show that the proposed algorithm performs comparably or better than batch methods, including the batch LRR, and significantly outperforms state-of-the-art online methods.
- Diversified Semantic Distribution Matching for Dataset Distillation27
A novel Diversified Semantic Distribution Matching (DSDM) approach for dataset distillation that matches the distilled data of the current pre-trained model with the historical integrated prototypes to accurately capture semantic features.
- 27
- MMSum: A Dataset for Multimodal Summarization and Thumbnail Generation of Videos26
This work has meticulously curated the MMSum dataset, providing a comprehensive dataset for this new direction of multimodal summarization with multimodal output, and released the MMSum dataset and the data collection tool as fully open-source resources.
- Asymmetric Deep Hashing for Efficient Hash Code Compression26
Extensive experiments on three widely used image retrieval benchmarks demonstrate that CCDH can significantly reduce the cost for compressing database codes when code length changes while keeping the state-of-the-art retrieval accuracy.
- A Hybrid Software Platform for Sony AIBO Robots26
A modular approach to the development of a hybrid soft- ware platform for Sony quadruped robots, which consists of an over- head vision system, a Sony AIBO robot and a desktop PC, and is designed to be interactive in order to speed up theDevelopment of various algorithms for ob- ject recognition and gait generation.
- 25
- 25
- A novel hierarchical attention-based triplet network with unsupervised domain adaptation for network intrusion detection25
HAT-UDA provides a novel approach for building a robust NIDS from benign traffic and available (known) attacks and proposes an unsupervised domain adaptation module in an adversarial manner to reduce the false positives of HAT- UDA when applied to new network scenarios.
- 25
- 24
- Context-Aware Recommendation via Graph-Based Contextual Modeling and Postfiltering24
A graph-based framework to model and incorporate contextual information into the recommendation process in an advantageous way and a probabilistic-based postfiltering strategy to refine the recommendation results as contextual conditions are explicitly given in a query are proposed.
- HotSnap: a hot distributed snapshot system for virtual machine cluster24
HotSnap is a VMC snapshot approach designed to enable taking hot distributed snapshot with milliseconds system downtime and TCP backoff duration, and only incurs tens of milliseconds in downtime and network interruption.
- 23
- Automatic Intrinsic Reward Shaping for Exploration in Deep Reinforcement Learning23
An intrinsic reward toolkit is developed to provide efficient and reliable implementations of diverse intrinsic reward approaches and demonstrates that AIRS can outperform the benchmarking schemes and achieve superior performance with simple architecture.
- 23
- CloudMon: a resource‐efficient IaaS cloud monitoring system based on networked intrusion detection system virtual appliances23
This work proposes a novel system, named CloudMon, which enables dynamic resource provision and live placement for NIDS‐VAs in IaaS cloud environments and verifies the effectiveness of the global resource scheduling approach by comparing it with two classic vector packing algorithms.
- Deep Uncoupled Discrete Hashing via Similarity Matrix Decomposition22
A novel asymmetric hashing method, called Deep Uncoupled Discrete Hashing (DUDH), for large-scale approximate nearest neighbor search, which exploits a small similarity-transfer image set to transfer the underlying semantic structures from the database to the query and implicitly keep the desired similarity.
- Multi-view Inference for Relation Extraction with Uncertain Knowledge22
This paper introduces ProBase, an uncertain KG that indicates to what extent a target entity belongs to a concept, and designs a novel multi-view inference framework to systematically integrate local context and global knowledge across three views: mention-, entity- and concept-view.
- 22
- Parallel Accessing Massive NetCDF Data Based on MapReduce22
The experiments show that the parallel method can be used to store and access the large scale NetCDF efficiently and is compared under a distributed environment based on PCs.
- 21
- UDora: A Unified Red Teaming Framework against LLM Agents by Dynamically Hijacking Their Own Reasoning21
UDora is presented, a unified red teaming framework designed for LLM agents that dynamically hijacks the agent's reasoning processes to compel malicious behavior and demonstrates superior effectiveness compared to existing methods across three LLM agent datasets.
- Co-saliency Detection Based on Hierarchical Consistency21
This work designs a novel superpixel-wise variational autoencoder network to precisely distinguish the salient objects from the background collection based on the reconstruction errors and proposes a two-stage clustering strategy to explore the multi-level salient objects consistency by using high-level and low-level features separately.
- 20
- 20
- 20
- BBDM: Image-to-image Translation with Brownian Bridge Diffusion Models19
A novel Vector Quantized Brownian Bridge (VQBB) diffusion model is proposed, which can model the transformation between two domains more accurate and accessible than the existing Markov diffusion methods.
- Binary Neural Network Hashing for Image Retrieval19
This paper constructs an efficient binarized network architecture to provide lighter model and faster inference, which directly generates binary outputs as the desired hash codes without introducing the quantization loss, and introduces a simple yet effective activation-aware loss to explicitly guide the updating of activations in intermediate layers.
- Multi-Feature Graph Attention Network for Cross-Modal Video-Text Retrieval19
A novel Multi-Feature Graph ATtention Network (MFGATN) is proposed, which enriches the representation of each feature in videos with the interchange of high-level semantic information among them and elaborately design a novel Dual Constraint Ranking Loss (DCRL), which simultaneously considers the inter-modal ranking constraint and the intra- modal structure constraint.
- 19
- 18
- 18
- 18
- Interpretable R-CNN18
The proposed AOGParsing operator aims to harness both the explainable rigor of top-down hierarchical and compositional grammar models and the discriminative power of bottom-up deep neural networks through end-to-end training.
- Benchmarking Robustness of Multimodal Image-Text Models under Distribution Shift17
It is observed that multimodal models are not robust to image and text perturbations, especially to image perturbations, so several new multimodal robustness benchmarks are proposed by applying 17 image perturbation and 16 text perturbation techniques on top of existing datasets.
- 17
- 17
- Design of redundant robot painting system for long non-regular duct17
A special automatic redundant robot painting system (RRPS), which can automatically navigate and paint in the long non-regular duct, designed with three subsystems: a redundant robot, a spraying system and a control and safety system.
- Optimizing annotation resources for natural language de-identification via a game theoretic framework17
A game theoretic framework can be applied in leading HCO’s to optimized decision making in natural language de-identification investments before sharing EMR data, enabling a more refined cost-benefit tradeoff, improving both privacy and utility for the HCO.
- 17
- 17
- 17
- Handling Label Uncertainty for Camera Incremental Person Re-Identification16
This paper investigates person re-identification in an unexplored scenario named Camera Incremental Person ReID (CIPR), which advances existing lifelong person ReID by taking into account the class overlap issue and proposes a novel framework ExtendOVA, which significantly outperforms the state-of-the-arts with remarkable advantages.
- 16
- Attack is the Best Defense: Towards Preemptive-Protection Person Re-Identification16
It is argued that directly applying adversary-based methods to protect the ReID system is sub-optimal due to the 'overlap identity' issue, and a novel preemptive-Protection person Re-IDentification (PRIDE) method is proposed, which explicitly constraining the adversarial image to an isolated location.
- Reducing the Computational Complexity of Two-Dimensional LSTMs16
This work finds that the fbGrid-LSTM can reduce computation costs by a factor of four with no loss in accuracy, on a 12,500 hour Voice Search task.
- 16
- Scalable Visual State Space Model with Fractal Scanning15
Fractal curves maintain high spatial proximity and adapt to different image resolutions, avoiding redundancy and enhancing SSMs' ability to model complex patterns accurately.
- 15
- 15
- 15
- Application Algorithms for Basketball Training Based on Big Data and Internet of Things15
The proposed algorithm uses gesture recognition from continuous video movement characteristics of key figures, trajectory characteristics, and background information fusion to improve the recognition mechanism based on the Shuangliu C3D video basketball player action classification method.
- 15
- Some characterizations and properties of COM-Poisson random variables15
Under the property of “not closed under addition” of COM-Poisson random variables, a new characterization of Poisson distribution is found and a new property of Stam inequality for COM-type discrete version Fisher information is obtained.
- 15
- 15
- Active appearance models using statistical characteristics of Gabor based texture representation15
A new AAM method based on Gabor texture feature representation is proposed, which is the first time to extract statistical features from both Gabor magnitude and Gabor phase as the texture representation in AAM.
- 15
- 14
- 14
- COLEP: Certifiably Robust Learning-Reasoning Conformal Prediction via Probabilistic Circuits14
This work proposes a certifiably robust learning-reasoning conformal prediction framework (COLEP) via probabilistic circuits, which comprise a data-driven learning component that trains statistical models to learn different semantic concepts, and a reasoning component that encodes knowledge and characterizes the relationships among the trained models for logic reasoning.
- 14
- Universal Domain Adaptive Network Embedding for Node Classification14
This work proposes a novel Universal Domain Adaptive Network Embedding (UDANE) framework, which learns transferable node representations across networks to succeed in node classification and demonstrates that the proposed UDANE model outperforms the state-of-the-art baselines by a large margin.
- Large vocabulary speech recognition for languages of Africa: multilingual modeling and self-supervised learning14
The results show that pooling the small amounts of data available in multilingual end-to-end models, and pre-training on unsupervised data can help improve speech recognition quality for many African languages.
- MHMS: Multimodal Hierarchical Multimedia Summarization14
This work proposes a multimodal hierarchical multimedia summarization framework by interacting visual and language domains to generate both video and textual summaries and formulates a cross-domain alignment objective with optimal transport distance.
- Global Convergence and Induced Kernels of Gradient-Based Meta-Learning with Neural Nets.14
It is proved that GBML is equivalent to a functional gradient descent operation that explicitly propagates experience from the past tasks to new ones and a new kernel-based meta-learning approach is developed that outperforms GBML with standard DNNs on the Omniglot dataset when the number of past tasks for meta-training is small.
- Weighted Lasso estimates for sparse logistic regression: non-asymptotic properties with measurement errors14
This work proposes two types of weighted Lasso estimates, depending upon covariates determined by the McDiarmid inequality, and compares the performance of the method with that of former weighted estimates on simulated data, then applies it to do real data analysis.
- SERAC3: Smart and economical resource allocation for big data clusters in community clouds14
SERAC3 is a system that can automatically extract representative workloads from incoming big data analysis jobs, smartly decide an optimal configuration for each job, and adjust its assigning strategy in a quasi-realtime mode to allocate resources smartly and economically for big data clusters in community clouds.
- First-Order Sensitivity Analysis for Hidden Neuron Selection in Layer-Wise Training of Networks14
A first-order sensitivity analysis method, called random balance designs (RBD), is employed to select redundant neurons in hidden layers and multiple experimental results demonstrate the generality and effectiveness of the proposed approach on structural learning of neural networks.
- 14
- Pairwise-Label-Based Deep Incremental Hashing with Simultaneous Code Expansion13
A unified deep hash framework that can simultaneously learn new classes and increase hash code capacity is presented, and a pairwise-label-based incremental similarity-preserving loss to optimize the new CNN model, which can incrementally preserve new similarities while maintaining the old ones is proposed.
- Privacy of Autonomous Vehicles: Risks, Protection Methods, and Future Directions13
A new taxonomy for privacy risks and protection methods in AVs is provided, and privacy in AVs is classified into three levels: individual, population, and proprietary, to help shape the privacy research in AV and guide the privacy protection technology design.
- 13
- 13
- 12
- VQRAE: Representation Quantization Autoencoders for Multimodal Understanding, Generation and Reconstruction12
VQRAE, a Vector Quantization version of Representation AutoEncoders, is proposed, which pioneers the first exploration in unified representation to produce Continuous semantic features for image understanding and Discrete tokens for visual generation within a unified tokenizer.
- 12
- ELSEIR: A Privacy-Preserving Large-Scale Image Retrieval Framework for Outsourced Data Sharing12
This paper designs an irreversible random hash code generation method that incorporates permutation keys for personalized access and integrates differential privacy to further enhance data security, and theoretically proven that the ELSEIR guarantees both outsourced data security and query user privacy.
- Semantics-Consistent Cross-domain Summarization via Optimal Transport Alignment12
A Semantics-Consistent Cross-domain Summarization (SCCS) model based on optimal transport alignment with visual and textual segmentation is proposed, which first decomposes both video and article into segments in order to capture the structural semantics and follows a cross-domain alignment objective with optimal transport distance.
- Rewarding Episodic Visitation Discrepancy for Exploration in Reinforcement Learning12
Rewarding Episodic Visitation Discrepancy (REVD), a computation-efficient and quantified exploration method that provides intrinsic rewards by evaluating the R\'enyi divergence-based visitation discrepancy between episodes, is proposed.
- FOCUS: Fairness via Agent-Awareness for Federated Learning on Heterogeneous Data12
This work proposes a formal FL fairness definition, fairness via agent-awareness (FAA), which takes different contributions of heterogeneous agents into account and shows that on four FL datasets, FOCUS achieves significantly higher fairness in terms of FAA while maintaining competitive prediction accuracy compared with FedAvg and state-of-the-art fair FL algorithms.
- 12
- Fast and Multilevel Semantic-Preserving Discrete Hashing.12
A novel deep asymmetric discrete hashing method, named F ast and M ultilevel semantic-preserving D iscrete H ashing(FMDH), which makes the best of supervised information to preserve the multilevel semantic similarities between multi-label images, and further accelerates the training process.
- Database Audit Workload Prioritization via Game Theory12
An auditing approach that accounts for adversarial behavior by prioritizing the order in which types of alerts are investigated and providing an upper bound on how much resource to allocate for each type is introduced.
- Get Your Workload in Order: Game Theoretic Prioritization of Database Auditing12
A novel approach to database auditing is introduced that explicitly accounts for adversarial behavior by prioritizing the order in which types of alerts are investigated and providing an upper bound on how much budget to allocate for auditing each alert type.
- 12
- 11
- SCCS: Semantics-Consistent Cross-domain Summarization via Optimal Transport Alignment11
This work proposes a Semantics-Consistent Cross-domain Summa-rization (SCCS) model based on optimal transport alignment with visual and textual segmentation, and follows a cross-domain alignment objective with optimal transport distance, which leverages multimodal interaction to match and select the visual and textual summary.
- Group Distributionally Robust Reinforcement Learning with Hierarchical Latent Variables11
Group Distributionally Robust Markov Decision Process (GDR-MDP), a flexible hierarchical MDP formulation that encodes task groups via a latent mixture model that identifies the optimal policy that maximizes the expected return under the worst-possible qualified belief over task groups within an ambiguity set is proposed.
- Sparse Mixture-of-Experts are Domain Generalizable Learners11
This work reveals the mixture-of-experts (MoE) model’s generalizability on DG by leveraging to distributively handle multiple aspects of the predictive features across domains by incorporating sparsity and fusion mechanisms into the MoE framework to keep the model both sparse and predictive.
- On the effectiveness of partial variance reduction in federated learning with heterogeneous data11
This paper revisits the widely used FedAvg algorithm in a deep neural network and proposes to correct model drift by variance reduction only on the final layers, which significantly outperforms existing benchmarks at a similar or lower communication cost.
- Efficient Hash Code Expansion by Recycling Old Bits11
This paper elaborately design an end-to-end asymmetric framework to simultaneously optimize a CNN model for query images and a code projection matrix for database images and proposes to adopt smooth similarity matrix to solve "similarity contradiction" problem existing in multi-label image datasets, thus further improving the performance on multi- label datasets.
- A Refined Method of High-Resolution Remote Sensing Change Detection Based on Machine Learning for Newly Constructed Building Areas11
It can be concluded that adoption of area registration improves registration efficiency, and the joint use of prior rules and SVM classifier with probability features could avoid over and missing detection for NCBAs.
- Projection-Wise Disentangling for Fair and Interpretable Representation Learning: Application to 3D Facial Shape Analysis11
This work proposes to mitigate the bias while keeping almost all information in the latent representations, which enables us to observe and interpret them as well, and enforce the independence between biases and projected features rather than all learned features.
- 11
- Deep Discrete Attention Guided Hashing for Face Image Retrieval11
This work proposes a deep hashing method specially designed for face image retrieval named deep Discrete Attention Guided Hashing (DAGH), where the discriminative power of hash codes is enhanced by a well-designed discrete identity loss.
- 11
- 11
- 11
- 11
- Scalable Iterative Classification for Sanitizing Large-Scale Datasets11
This work model data sanitization as a game between a publisher who chooses a set of classifiers to apply to data and publishes only instances predicted as non-sensitive and an attacker who combines machine learning and manual inspection to uncover leaked identifying information.
- A Novel Naive Bayes Classification Algorithm Based on Particle Swarm Optimization11
This paper presents an approach called particle swarm optimization-naive Bayes (PSO-NB) which takes advantage of combination particle Swarm optimization with naive Bayes for attribute selection to improve naive Baye classifier.
- Precomputed Clustering for Movie Recommendation System in Real Time11
A novel idea that applies machine learning techniques to construct a cluster for the movie by implementing a distance matrix based on the movie features and then make movie recommendation in real time is presented.
- Improved probability hypothesis density filter for multitarget tracking11
An improved PHD filter that can achieve MTT with better performance by applying particle approximation to achieve the closed-form solution to the PHD propagations and simplified by the method of kernel estimate.
- EAIRA: Establishing a Methodology for Evaluating AI Models as Scientific Research Assistants10
A multifaceted methodology for Evaluating AI models as scientific Research Assistants (EAIRA) developed at Argonne National Laboratory is described, designed to be generalizable to a wide range of scientific domains.
- 10
- 10
- Wse-MF: A weighting-based student exercise matrix factorization model10
A novel approach called Weighting-based Student Exercise Matrix Factorization (Wse-MF) which combines student learning ability and exercise difficulty as prior weights is presented, and improvements to matrix factorization within the context of educational data are demonstrated.
- Is Vertical Logistic Regression Privacy-Preserving? A Comprehensive Privacy Analysis and Beyond10
It is suggested that all vertical federated learning frameworks that solely depend on HE might contain severe privacy risks, and DP, which has already demonstrated its power in horizontal Federated learning, can also play a crucial role in the vertical setting, especially when coupled with HE or secure multi-party computation techniques.
- 10
- 10
- A Unified Framework for Anomaly Detection of Satellite Images Based on Well-Designed Features and an Artificial Neural Network10
This study presents a unified framework based on well-designed features an Artificial Neural Network (ANN) to automatically identify defective images and an effective method was proposed to explain the inference of ANN based on local gradient approximation.
- Neural Zero-Inflated Quality Estimation Model for Automatic Speech Recognition System10
A novel neural zero-inflated model is proposed to predict the WER of the ASR result without transcripts and achieves better performance on WER prediction in the metrics of Pearson and MAE, compared with most existed quality estimation algorithms for ASR or machine translation.
- Meta-Graph Based Attention-Aware Recommendation over Heterogeneous Information Networks10
A Meta-Graph based Attention-aware Recommendation (MGAR) over HINs, which utilizes rich meta-graph based latent features to guide the heterogeneous information fusion recommendation and proposes an attention-based feature enhancement model which enables useful features and useless features contribute differently to the prediction, thus improves the performance of the recommendation.
- Multi-scale Quantum Harmonic Oscillator Algorithm with Individual Stabilization Strategy10
A MQHOA with individual stabilization strategy (IS-MQHOA) is proposed utilizing the individual steady criterion instead of the group statistics, which is more rigorous for the particles in the energy level stabilization process.
- Behavioral Experiments in Email Filter Evasion10
A synthetic model of human evasion behavior is developed which closely matches observed behavior and effectively replicates experimental findings in simulation and suggests that aggressive feature reduction — a common practice in applied machine learning — can actually facilitate evasion.
- 10
- Dawning Nebulae: A PetaFLOPS Supercomputer with a Heterogeneous Structure10
System tuning methodologies aiming at petaFLOPS Linpack result are presented, including algorithmic optimization and communication improvement and the design of its file I/O subsystem, including HVFS and the underlying DCFS3 is described.
- 10
- 9
- 9
- 9
- Django: Bilateral coflow scheduling with predictive concurrent connections9
This paper first uses Support Vector Machine (SVM) as the machine learning model to automatically identify the optimal number of concurrent connections, i.e., the queue limitation in the switch, and develops a set of distributed coflow scheduling algorithms in a scalable manner.
- Flexible and Fast Spatial Return Level Estimation Via a Spatially Fused Penalty9
A flexible and fast approach based on the spatial GEV and spatial GPD is proposed by introducing fused lasso and fused ridge penalty for parameter regularization that enables improved return level estimation for large spatial extremes compared to the existing methods.
- 9
- Community Focusing: Yet Another Query-Dependent Community Detection9
A new problem called community focusing (CF) is studied, which finds a community where the members are close and densely connected to the query nodes, and effective methods are devised for CF.
- Web Application-Layer DDoS Attack Detection Based on Generalized Jaccard Similarity and Information Entropy9
An application-layer DDoS attack detection system based on user behavior anomaly detection is constructed and test results indicate that the system can describe user behavior well and detect DDoS attacks effectively.
- 9
- Self‐normalization for Spatial Data9
Inference for both spatial lattice data with possibly irregularly shaped sampling region and non‐lattice data is considered, by extending the recently proposed self‐normalization (SN) approach from stationary time series to the spatial setup.
- 9
- A Speech Enhancement Method Based on Kalman Filtering9
This study presents a monaural speech enhancement method based on spectral subtraction and Kalman filtering by extracting the Liljencrants–Fant excitation during voiced speech, in which the nature of glottal flow can be maintained and could achieve significant improvements on objective quality.
- 8
- 8
- An Interpretable Machine Learning Model with Deep Learning-based Imaging Biomarkers for Diagnosis of Alzheimer's Disease8
The proposed framework significantly outperformed an EBM model using volume biomarkers instead of deep learning-based features, as well as an end-to-end convolutional neural network (CNN) with optimized architecture.
- 8
- 8
- Efficient Skyline Query over Multiple Relations8
An efficient skyline join algorithm Skyjog is proposed, which is applicable for query on two or more relations, and outperforms the state-of-the-art skyline join algorithms on two and more than two relations.
- 7
- 7
- Camera-specific Informative Data Augmentation Module for Unbalanced Person Re-identification7
A novel camera-specific informative data augmentation module~(CIDAM) to alleviate the proposed camera-unbalanced problem and can be compatible with a variety of existing Re-ID methods.
- Flexible Order Aware Sequential Recommendation7
A flexible order aware Sequential Recommendation (FOSR) method to model dependencies from flexible orders comprehensively and learn dynamic user interests effectively is proposed, which outperforms various state-of-the-art sequential recommendation methods.
- Rescuing Deep Hashing from Dead Bits Problem7
This paper proposes a simple but effective gradient amplifier which acts before activation functions to alleviateDBP, and devise an error-aware quantization loss to further alleviate DBP.
- 7
- Cross-Modal Interaction-Aware Progressive Fusion Network for Drug–Target Interaction Prediction6
A deep learning fusion framework termed cross-modal interaction-aware progressive fusion network (CIPFN) for DTI prediction is proposed, which introduces a bidirectional interaction-aware module to precisely align fine-grained interactions between drugs and proteins.
- CIRAG: Retrieval-Augmented Language Model with Collective Intelligence6
CIRAG first enhances retrieval diversity by expanding queries based on extracted entities, then combines frequency-based and semantic-based reranking to form a multi granularity fusion reranking thereby assessing better relevance, and integrate multiple information sources for accurate content generation.
- Distribution Network Anomaly Detection Based on Graph Contrastive Learning6
DNT-GCL is presented, a novel few-shot anomaly detection technique based on graph contrastive learning that achieves an accuracy improvement of at least 8.75% over baseline methods and demonstrates remarkable performance in few-shot scenarios.
- 6
- 6
- An Efficient Approach to Manage Natural Noises in Recommender Systems6
A new approach to managing natural noises in recommendation systems by providing the detection criteria for natural noises based on the classifications of users and items and correcting them with threshold values weighted by probabilities is presented.
- A Diagnostic Model of Volleyball Techniques and Tactics Based on Wireless Communication Network6
The method of LTE system and MIMO technology is proposed and the tactics analysis model based on the characteristics of the wireless network is designed, which combined volleyball tactics and game theory to design a wireless communication network volleyball tacticsAnalysis model.
- Heterogeneous Side Information-based Iterative Guidance Model for Recommendation6
The experimental results show that ILIG outperforms the state-of-the-art recommender systems and captures the interrelation among various types of heterogeneous side information to comprehensively learn the representations of users and items.
- 6
- 6
- 6
- 6
- Robust Determinantal Generative Classifier for Noisy Labels and Adversarial Attacks6
This paper proposes a novel inference method, Deep Determinantal Generative Classifier (DDGC), which can obtain a more robust decision boundary under any softmax neural classifier pre-trained on noisy datasets, and demonstrates the superiority of DDGC given different learning models optimized by various training techniques to handle noisy labels or adversarial samples.
- 6
- Multi-view VR system for co-located multidisciplinary collaboration and its application in ergonomic design6
Results show that using multi-view VR system achieves a better effectiveness and user performance and an application for collaborative ergonomic design is explored and an industrial multidisciplinary scenario is realized with proposed system.
- 6
- Surface skeleton extraction and its application for data storage in 3D sensor networks6
This paper proposes a scalable and distributed connectivity-based algorithm to extract the surface skeleton of 3D sensor networks and proposes a novel approach to identifying surface skeleton nodes by computing the extended feature nodes such that it is robust against boundary noise, etc.
- 6
- 5
- Uneven Event Modeling for Partially Relevant Video Retrieval5
An Uneven Event Modeling (UEM) framework for PRVR is proposed, which first introduces the Progressive-Grouped Video Segmentation (PGVS) module, and proposes the Context-Aware Event Refinement (CAER) module to refine the event representation conditioned the text’s cross-attention.
- ARMs: Adaptive Red-Teaming Agent against Multimodal Models with Plug-and-Play Attacks5
It is shown that the diversity of red-teaming instances generated by ARMs is significantly higher, revealing emerging vulnerabilities in VLMs, and construct ARMs-Bench, a large-scale multimodal safety dataset comprising over 30K red-teaming instances spanning 51 diverse risk categories, grounded in both real-world multimodal threats and regulatory risks.
- 5
- 5
- 5
- Effective and Efficient Federated Tree Learning on Hybrid Data5
This work proposes HybridTree, a novel federated learning approach that enables federated tree learning on hybrid data and proposes a layer-level solution that does not need frequent communication traffic to train a tree.
- Your Autoregressive Generative Model Can be Better If You Treat It as an Energy-Based One5
This paper proposes a unique method termed E-ARM for training autoregressive generative models that takes advantage of a well-designed energy-based learning objective and can be trained efficiently and is capable of alleviating the exposure bias problem and increase temporal coherence for autore Progressive Generative models.
- 5
- Online Markov Blanket Learning for High-Dimensional Data5
Two online MB learning algorithms are proposed, called O nline S imul T aneous MB learning (O-ST) algorithm and O-DC algorithm, which achieve higher efficiency and better accuracy than the state-of-the-art onlineMB learning algorithms.
- 5
- 5
- 5
- A Novel Classification Scheme of Moving Targets at Sea Based on Ward's and K-means Clustering5
The simulation results show that the proposed scheme is applicable to classify and identify the moving targets at sea.
- Neural Relation Classification Using Selective Attention and Symmetrical Directional Instances5
This work presents a novel CNN-based sentence encoder with selective attention by leveraging the shortest dependency paths, and devise a classification framework using symmetrical directional—forward and backward—instances via information fusion.
- Object Detection via End-to-End Integration of Aspect Ratio and Context Aware Part-based Models and Fully Convolutional Networks.5
The integrated framework is fully convolutional and enjoys end-to-end training, which is called the aspect ratio and context aware fully Convolutional network (ARC-FCN) and shows very competitive results on the PASCAL VOC datasets.
- 5
- 5
- Fine-Tuning Jailbreaks under Highly Constrained Black-Box Settings: A Three-Pronged Approach4
A three-pronged jailbreak attack that combines safety-styled prefix/suffix wrappers, benign lexical encodings of sensitive tokens, and a backdoor mechanism, enabling the model to learn harmful behaviors while individual datapoints appear innocuous.
- 4
- Heterogeneous Data Game: Characterizing the Model Competition Across Multiple Data Sources4
This paper proposes a game-theoretic framework -- the Heterogeneous Data Game -- to analyze how providers compete across heterogeneous data sources, investigating the resulting pure Nash equilibria (PNE) and offering theoretical insights into both homogeneous and heterogeneous PNEs.
- 4
- 4
- Remote intelligent identification of suspended foreign objects in substation inspection images based on edge optimization faster R-CNN algorithm4
The experimental results demonstrate that this substation inspection method possesses a robust capability to enhance substation inspection images, can effectively remotely identify various types of foreign objects hanging in these images, provides valuable data on foreign object hangings for substation operations and arcs, and exhibits strong applicability.
- A method of combining traffic classification and traffic prediction based on machine learning in wireless networks4
The experimental results show that the proposed method of joint wireless network traffic classification and traffic prediction based on machine learning improves both the accuracy of traffic classificationand traffic prediction in wireless networks.
- 4
- 4
- Diversifying Emotional Dialogue Generation via Selective Adversarial Training4
This work proposes a method to generate diverse emotional responses through selective perturbation, and shows that this model outperforms existing baselines in terms of emotional expression and response diversity.
- Learning Pair-Centric Representation for Link Sign Prediction with Subgraph4
A new perspective to conduct link sign prediction within the paradigm of subgraph classification is provided and a novel Subgraph-based link Sign Prediction (SSP) model is proposed that consistently and significantly outperforms all the state-of-the-art baselines.
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- Affective feature knowledge interaction for empathetic conversation generation4
A novel affective feature knowledge interactive model named AFKI is proposed, to enhance response generation performance, which enriches conversation history to obtain emotional interactive context by leveraging fine-grained emotional features and commonsense knowledge.
- 4
- 4
- 4
- Beautify As You Like4
The first fast makeup transfer system named as Fast Pose and expression robust Spatial-Aware GAN (FPSGAN) is presented, which is robust to face pose and expression and light-weighted and fast.
- 4
- High-Performance OCR on Packing Boxes in Industry Based on Deep Learning4
In this work, a high-performance OCR method based on deep learning is proposed, which only needs 100 training samples with the size of 640 * 480 to achieve an accuracy of 99.92%, and the whole detection process requires only about 78 ms per image on average.
- Joint Learning from Labeled and Unlabeled Data for Information Retrieval4
It is hypothesized that general semantics learned from unlabeled data can complement task-specific representation learned from labeled data of limited quality, and that a combination of the two is favorable.
- Efficient Algorithms of Parallel Skyline Join over Data Streams4
A Novel Iterative framework is developed based on the existing work and the inherent parallelism of the Novel Iteratives is studied, and two parallel skyline join algorithms over sliding windows, NP-SWJ and IP- SWJ are proposed.
- 4
- 4
- A Dynamic Nonparametric Model for Characterizing the Topical Communities in Social Streams4
This work proposes a fully nonparametric Dynamic Topical Community Model (nDTCM), where infinite latent community variables coupled with infinite latent topic variables in each epoch, and the temporal dependencies between variables across epochs are modeled via the rich-gets-richer scheme.
- 4
- Feature extraction for identification of drugs and explosives concealed by body packing based on discrete cosine transform plus linear discriminant analysis4
DCT plus LDA is effective in featured extraction for identification of drugs and explosives concealed by body packing and the recognition rates obtained are more accurate than expected.
- Least Squares Constant Modulus Blind Adaptive Beamforming with Sparse Constraint4
Simulation results indicate that the proposed approach exhibits better performance than the well-known least squares constant modulus algorithm (LSCMA) in beamforming with sparse constraint.
- 4
- A DATA COLLECTION MODEL FOR INTRUSION DETECTION SYSTEM BASED ON SIMPLE RANDOM SAMPLING4
Experimental results show the new data collection model for IDS is able to improve the efficiency of data collection and strengthen the processing performance of IDS with little devaluation of detection precision.
- 4
- 3
- DC-MSSFF Net: Dule-channel multi-scale spatial-spectral feature fusion network for cholangiocarcinoma pathology high-resolution hyperspectral image segmentation3
The superior performance of the DC-MSSFF network pioneers the inductive learning task of deep frameworks for semantic segmentation of high-resolution hyperspectral image (HR-HSI).
- 3
- 3
- 3
- Targeted Transferable Attack against Deep Hashing Retrieval3
This paper proposes a new Iterative Anchor code Optimization method to generate anchor code with superior representative semantics of target label, which can improve both targeted white-box and black-box performances and proposes a novel Targeted Transferable Attack method against deep hashing with Generative Adversarial Network (TTA-GAN).
- 3
- Pose-Controllable 3D Facial Animation Synthesis using Hierarchical Audio-Vertex Attention3
A novel pose-controllable 3D facial animation synthesis method is proposed by utilizing hierarchical audio-vertex attention by utilizing the 2D talking face technique and results indicate that the proposed method can produce more realistic facial expressions and head posture movements.
- 3
- 3
- Generative adversarial networks based on MLP3
This paper extends MLP to the task of image generation based on generative adversarial network (GAN) and uses a hybrid model to solve the problem which uses MLP blocks as generator and CNN blocks as discriminator.
- 3
- 3
- 3
- 3
- 3
- Research on English Translation Based on Recursive Deep Neural Network3
The method of this model is superior to the existing system in terms of cross-entropy loss measurement and the performance of the GRU and LSTM layers is better than that of LSTM.
- 3
- Marginalized Graph Attention Hashing for Zero-Shot Image Retrieval3
A novel deep zero-shot hashing method, named MGAH, which introduces the masked attention mechanism to construct a joint-semantics similarity graph, which captures the intrinsic relationship from different metric spaces, making it competent to transfer knowledge from seen classes into unseen classes.
- Discriminative Deep Attention-Aware Hashing for Face Image Retrieval3
An end-to-end deep hashing method with attention mechanisms to learn discriminative hash codes to enhance the discrimination of face features from the spatial aspect and an attention-aware hash network is designed to learn compact and discriminatives hash codes.
- 3
- Skyline Join Query Processing over Multiple Relations3
An efficient algorithm Skyjog is proposed, which is applicable for skyline join on two or even more relations, and has an outstanding performance on all datasets, and outperforms the state-of-the-art skyline join algorithms on both two relations and more than two relations.
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- Dangerous Language Habits! Exploiting Code-Mixing for Backdoor Attacks on NLP Models2
This work proposes a novel backdoor attack strategy that leverages the linguistic properties of code-mixing, a language form that combines elements from two or more languages, as implicit triggers for text backdoor attacks, and designs three types of linguistically grounded triggers.
- Robust Leakage Detection in Tunnels with Unstable Illumination Using an Explainable Deep Learning Approach2
Visual explanations revealed that enhancing low-level feature extraction was critical for encoder optimization, and GAU significantly improved feature connections, and a deeper decoder stabilized feature extraction, resulting in a more stable and interpretable training process.
- ShieldIR: Privacy-Preserving Unsupervised Cross-Domain Image Retrieval via Dual Protection Transformation2
ShieldIR unifies intra-domain and cross-domain representation learning through a Dual Protection Transformation (DPT) module, which introduces a structured feature space via orthogonal projection and ensures data privacy by adding calibrated differential privacy noise.
- Two-Stage Adversarial Training for Deep Hashing via Representation Distillation2
A novel Clean model Representation Distillation based Adversarial Training (CRDAT) method, which enables the robust model to learn both discriminative semantic information and robust patterns by separating these two losses into two stages, i.e., standard training stage of a clean teacher model and adversarial training stage of a robust student model.
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- Group Lasso based redundancy-controlled feature selection for fuzzy neural network2
A group Lasso based redundancy-controlled feature selection (input pruning) method to optimize the structure of the fuzzy rule based neural network and demonstrates the effectiveness of the method on two benchmark classification datasets.
- Efficient Associate Rules Mining Based on Topology for Items of Transactional Data2
This paper deduces a binary relation on itemset, and builds a topology space of itemset based on the binary relation and the quotient lattice of the topology according to transactions of itemsets, and provides efficient algorithms to generate Min-Max association rules or reduce generalized association rules based onThe lower approximation and the upper approximation of a template, respectively.
- Invariant-Feature Subspace Recovery: A New Class of Provable Domain Generalization Algorithms2
Invariant-feature Subspace Recovery (ISR) is proposed: a new class of algorithms to achieve provable domain generalization across the settings of classification and regression problems and, unlike IRM, bypass non-convexity issues and enjoy global convergence guarantees.
- 2
- 2
- 2
- 2
- 2
- 2
- Research of Vertical Domain Entity Linking Method Fusing Bert-Binary2
To solve the problem of unclear entity boundaries and low recognition accuracy in Chinese text, the crop dataset is constructed and a Bert-binary-based entity link method is proposed and the experimental results show the effectiveness of this method.
- 2
- 2
- 2
- Research and application of patent value rapid screening algorithm based on big data and neural network2
The method proposed in this paper can stimulate the potential value of patents and efficiently promote the commercialization and product transformation of patent achievements and has reference significance for improving the transformation efficiency of scientific and technological achievements.
- 2
- 2
- 2
- 2
- Unsupervised Cross-Modal Retrieval by Coupled Dual Generative Adversarial Networks2
This paper addresses the unsupervised cross-modal retrieval problem using a novel framework called coupled dual generative adversarial networks (CDGAN), which can well match images and sentences with complex content, and it can achieve the state-of-the-art cross- modal retrieval results on two popular benchmark datasets.
- Large-scale Agent Data Partitioning Based on DensityRepel-K_medoids2
An improved K_medoids algorithm (DensityRepel-K_medoid), which is implemented by high-performance programming language X10 and applied to large-scale Agent data partitioning simulation based on distance interaction, and avoids the defects of K_means clustering algorithm sensitive to outliers.
- 2
- 2
- 2
- Enhanced Query Classification with Millions of Fine-Grained Topics2
Empirical evaluation confirms that the topic based classification algorithms can significantly enhance the performances of traditional classifiers in read-world query classification tasks.
- 2
- Extending PowerPack for Profiling and Analysis of High-Performance Accelerator-Based Systems2
Detailed experimental studies of the power usage and distribution of Xeon-Phi-based systems in comparison to the NVIDIA Tesla and an Intel Sandy Bridge multicore host processor help explain the causes of power-performance scalability for a set of HPC applications.
- A Layered Context-Aware Agent for Mobile Applications Based on Users Needs Hierarchy2
An experiment was conducted on the comparison between two methods of the random model and the proposed layered model for sending recommendation messages and the result verified that the proposed agent was effective.
- 2
- 2
- 2
- Essex Rovers 2001 Team Description2
The research efforts to build the Essex Rovers'01 robot soccer team participated in the RoboCup-2001 competition are introduced, which consists of three modules such as Perception module, Cognition module and Action module that are used for the team to achieve intelligent actions in real time.
- Plasma Shape Control via Zero-shot Generative Reinforcement Learning1
This paper proposes a novel framework for developing a versatile, zero-shot control policy from a large-scale offline dataset of historical PID-controlled discharges, which synergistically combines Generative Adversarial Imitation Learning (GAIL) with Hilbert space representation learning to achieve dual objectives: mimicking the stable operational style of the PID data and constructing a geometrically structured latent space for efficient, goal-directed control.
- 1
- Retrieval-Augmented Image Captioning via Synthesized Entity-Aware Knowledge Representations1
SEAR is proposed, a novel framework that utilizes external Synthesized Entity-Aware knowledge Representations to improve captioning performance, outperforming existing lightweight RAIC methods and remaining competitive with heavyweight models.
- 1
- 1
- Learning double balancing representation for heterogeneous dose–response curve estimation1
This paper theoretically demonstrate the importance of the balancing and prognostic representations for unbiased estimation of the heterogeneous dose-response curves and proposes an end-to-end Contrastive balancing Representation learning Network (CRNet) and a three-stage Weighted Double Balancing Network (WDBN) using a partial distance measure.
- 1
- 1
- Video anomaly detection via improved future frame prediction1
This work proposes a light but efficient framework that builds upon future frame prediction techniques and incorporates Convolutional Long Short-Term Memory, masked convolution, and attention mechanisms to enhance the detection accuracy and simplify the model's complexity.
- 1
- 1
- 1
- Solving the Performance Issues of Epsilon Estimation Method in Differentially Private ERM: Analysis, Solution and Evaluation1
This paper comprehensively examines the performance issues of Li's method, identifies their underlying cause, and presents effective solutions that can greatly improve the performance of Li’s method in a highly efficient manner.
- 1
- 1
- 1
- 1
- Anti-interference Technology of Multi-type Data Acquisition and Transmission Based on HPLC1
It is concluded that suitable filters and amplifiers should be selected to improve the signal-to-noise ratio and the transmission quality of HPLC signals and an anti-interference technology for data transmission based on HPLC is proposed.
- 1
- 1
- 1
- 1
- 1
- 1
- Third Workshop on Adversarial Learning Methods for Machine Learning and Data Mining (AdvML 2021)1
The AdvML'21 workshop consists of three tracks: (i) open-call paper submissions; (ii) invited speakers; and (iii) rising star awards and presentations.
- 1
- An Adaptive Hierarchical Hybrid Multicast Based on Information-Centric Networking1
Simulation results show that compared with Source-Specific Multicast (SSM) and Bit Index Explicit Replication (BIER), AHHM can not only reduce the multicast forwarding states but also reduce the control overhead and link load.
- 1
- Secure Learning In Adversarial Environments1
The sensing-reasoning certifiable machine learning pipeline is introduced which aims to improve the certifiable robustness of ML by integrating domain knowledge expressed as logic rules and leveraging game theoretic analysis to proactively obtain domain knowledge and further improve the robustness of ML systems.
- 1
- Invariant Adversarial Learning for Distributional Robustness.1
This paper proposes the Invariant Adversarial Learning (IAL) algorithm that leverages heterogeneous data sources to construct a more practical uncertainty set and conduct robustness optimization, where covariates are differentiated according to the stability of their correlations with the target.
- 1
- Asymmetric Deep Cross-modal Hashing1
An Asymmetric Deep Cross-modal Hashing (ADCH) method is proposed to perform more effective hash learning by simultaneously preserving the semantic similarity and the underlying data structures.
- 1
- 1
- Robust and Index-Compatible Deep Hashing for Accurate and Fast Image Retrieval1
This paper proposes a novel Robust and Index-Compatible Deep Hashing (RICH) method to learn compact similarity-preserving binary codes, which focuses on improving the retrieval accuracy and time efficiency simultaneously.
- 1
- 1
- Neural Retrieval with Partially Shared Embedding Spaces1
It is argued that queries and documents should be mapped into different but overlapping embedding spaces, which is named Partially Shared Embedding Space (PSES) model, which consists of twoembedding spaces respectively for queries and Documents, and a shared embedding space capturing common features of two sources.
- Data Quality Evaluation: Methodology and Key Factors1
Data Quality Evaluation is becoming an institutionalized stage in data quality lifecycle and more and more practice is promoted by data management and user organization in specific fields especially in better informationalized application circumstance.
- Analysis on the Network Structure of Public Opinion transmission about social hot issues based on the microblogging platform1
This paper presents a method about an analysis on the network structure of public opinion transmission about social hot issue based on the microblogging platform and uses social network analysis methods, web crawler to get data, and “Ucinet” to calculate the central, intermediary and other data of the networkructure.
- 1
- 1
- Research on the Integration and Verification of Foundational Software and Hardware1
On the basis of analyzing the whole process of integration and verification of foundational SW&HW, the technology system was advanced and the process, the content and the flow were provided, and the environment of Integration and verification was constructed.
- Optimizing Service Selection Using Hybrid Multi-objective Genetic Algorithms1
An important measure of the web service, quality of service (QoS), determines the success of the combined services to some extent, highlighting the need for service users and providers to reach an agreement on QoS of web services.
- A SaaS-based Software Modeling for Bank Intermediary Business1
This paper focuses on establishing the SAAS-based BIB performance and benchmark architecture and proposes the SaaS-by-SaaS BIB Database Model, which is proved to be better than that in three-tier C/S architecture.
- LuBase: A Search-Efficient Hybrid Storage System for Massive Text Data1
Experimental results demonstrate that LuBase can reap huge fruits from Lucene index technique and bring significant performance improvement for Impala when querying HBase.
- A Security Risk Analysis Method for Information System Based on Information Entropy1
An information system security risk analysis method based on information entropy to measure the risk of information systems, introduce the information entropy theory, and get the value of therisk of various risk factors with the combination of qualitative analysis and quantitative calculation is proposed.
- 1
- Modeling Business for MDA Software Paradigm1
In this paper, MDA is extended with a business model, and corresponding theory and presentation of the business model are explored, especially the dependencies and transformation mechanism of its sub-models are discussed.
- Design and Implementation of IoT Security System Towards Campus Safety1
A display method based on the massive data storage and browser is presented and a campus network security platform and the prototype system are designed and developed and experiment is carried out to prove the feasibility of campus networkSecurity platform in logic and performance.
- 1
- 1
- Pushing edge detection to the limit : towards building semantic features for human emotion recognition1
Pushing edge detection to the limit : towards building semantic features for human emotion recognition in edge detection.
- Survey-based Study of Novel Visual Cryptography Mechanisms1
Novel VC techniques are introduced, which flexibly provide the progressive version and scalable mechanism for sharing secret image securely and can guarantee that no one can have access to any clues of the secret image unless a pre-defined number of shadows is collected.
- 1
- Research on Mining Classification Rules of Spare Parts Based on Grey Rough Set Theory1
Case study from the classification management of the spare parts of a maintenance factory shows that the proposed mining classification rules model of spare parts based on grey rough set theory can reduce the unnecessary attributes of spare part without affecting the results of classification.
- A Cryptographic Alternative for Preserving PHI in Compliance with HIPAA Privacy/Security Regulations1
Using cryptographic techniques, a novel model which can confirm essentials in HIPAA is provided, which can appropriately handle emergency cases and the correctness of authentication is confirmed under the BAN logic model.
- A Novel Local Sensitive Frontier Analysis for Feature Extraction1
In this paper, an efficient feature extraction method, named local sensitive frontier analysis (LSFA), is proposed, which tries to find instances near the crossing of the multi-manifold, which are sensitive to classification, to form the frontier automatically.
- Detecting TCP-based applications using packet size distributions1
This Thesis suggests that for those TCP-based network applications that do not use the Nagle Algorithm, the detection mechanism, which had been proved to be successful for UDP-based applications, could be also adopted without any modification.
- 1
- –
- –
- –
- –
- –
- RED: Robust Event-Guided Motion Deblurring with Modality-Specific Disentanglement–
A Robust Event-guided Deblurring (RED) network with modality-specific disentangled representation is proposed with a Robustness-Oriented Perturbation Strategy (RPS) that applies random masking to events, which exposes RED to incomplete patterns and then foster robustness against various unknown scenario conditions.
- –
- –
- Super-Resolution Correction Method for Substation Infrared Image Non-uniformity Based on Adaptive Learning Rate CNN–
Experiments show that the proposed super-resolution non-uniformity correction method for substation infrared images based on a convolutional neural network with an adaptive learning rate can effectively extract non-uniformity features from substation infrared images with improved feature extraction performance.
- –
- –
- Modeling and Simulation of Public Opinion Evolution Based on the SIS-FJ Model with a Bidirectional Coupling Mechanism–
The SIS-FJ model is proposed, which combines the strengths of the SIS (Susceptible–Infected–Susceptible) model in information dissemination and the FJ (Friedkin–Johnsen) model in opinion interaction, and considers individual heterogeneity from multiple aspects, including infection rate, recovery rate, and individual susceptibility.
- –
- –
- –
- Pointer meter recognition algorithm based on two-dimensional convolution and calculus accumulation–
An instrument reading recognition method based on two-dimensional convolution and calculus accumulation and a weighted multi-feature matching algorithm are proposed to improve the success rate of template matching and accurately obtain the center coordinates of the rotating shaft.
- –
- –
- –
- –
- TablePuppet: A Generic Framework for Relational Federated Learning–
This paper proposes TablePuppet, a generic framework for RFL that decomposes the learning process into two steps: learning over join (LoJ) followed by learning over union (LoU), and shows that TablePuppet achieves model accuracy comparable to the centralized baselines running directly atop the SQL results.
- –
- –
- –
- –
- –
- –
- Model Explainability in Physiological and Healthcare-based Neural Networks–
The proposed method can provide a more efficient and accurate way to monitor SpO2 using videos captured from consumer-grade smartphones, which can be especially useful in telehealth and health screening settings.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- x4 Super-Resolution Analysis of Magnetic Resonance Imaging based on Generative Adversarial Network without Supervised Images–
This paper constructs a data set close to the natural low-high resolution image pair through degenerate kernel estimation and noise injection and constructs a two-layer generated countermeasure network based on the design ideas of ESRGAN, PatchGAN, and VGG-19 to ensure the pixel quality of the MRI image required for medical diagnosis.
- –
- Small Target Detection Algorithm Based on LBP Model–
The Local Binary Pattern (LBP) model is introduced that describes the image texture characteristics, combines the LBP model to carry out a small target detection algorithm analysis and research, and analyzes the experimental results.
- –
- –
- –
- MUSE: Illustrating Textual Attributes by Portrait Generation.–
A novel approach, MUSE, to illustrate textual attributes visually via portrait generation by extending an image-to-image generative model to accept textual attributes and proposes a new attribute reconstruction metric to evaluate whether the generated portraits preserve the subject's attributes.
- –
- –
- –
- –
- A scalable real-time many-shadowed-light rendering system–
A new shadow rendering system, with a number of novel design to support large amount of shadowed lights in a large virtual environment, with real-time performance on mainstream GPUs is presented.
- –
- –
- Teaching Variability in a Core Systems Course–
A web-based system that allows students to submit their own systems-level code to a specialized cluster which then benchmarks it while systematically varying a number of ordinal and categorical variables that reflect environmental factors that can influence the performance of complex systems.
- –
- –
- Study of efficiency of multi view system in multi-disciplinary collaboration task–
The context of multi-disciplinary collaboration and the usage of multiple computer-human interfaces (chis) with domain specified software is introduced and the co-located multi-view system with the technology ofMulti-representation, multi interaction and virtual reality is proposed.
- Agent Based Simulation on Task Workflow Oriented Self-Organization of Cellular Organizational Network–
Experimental results shows the difference of the organizational performance according to different task requirements, which exhibits potential application in task workflow oriented organization design and optimization.
- –
- Tracking millions of query intents with StarrySky and its applications–
It is shown how real-world applications can benefit from query intent tracking results using StarrySky, a practical system for identifying and inferring millions of fine-grained query intents in daily sponsored search with high precision.
- –
- Editorial for Special Issue on Social Computing–
This special issue features six selected high-quality papers covering a variety of aspects related to social computing, including service provision, resource allocation and management, network control, security and privacy issues.
- –
- Skyline Query on Anti-correlated Distributions: From the Perspective of Spatial Index–
An efficient algorithm RSAC for anti-correlated distributions is proposed, which solely uses one index structure to implement such two steps, and achieves up to 3 times performance improvement than SOAD on various datasets.
- –
- –
- Improved Adaptive α-β Filtering Algorithm–
To predict dynamic targets by using tracking gains, an improved adaptive α-β filtering algorithm is presented and the pseudo acceleration is introduced to build generalized model for uniform linear and acceleration motions.
- Blur kernel estimate in single noisy image deblurring–
This work estimates the blur kernel accurately by applying a series of directional low-pass filters in different orientations to the input blurred image, and effectively constructing the Radon transform of the blur Kernel from each filtered image, then uses a robust non-blind deconvolution method with outlier handling to generate the final results.
- –
- A new metric method-improved structural holes researches on software networks–
A new and better method to metric software networks structural holes is proposed and shows that the new method can better reflect bridge role of vertexes on software network and there is a significant correlation between degree and improved structural holes.
- Improved Algorithm of Complex Spatial Co-Location Pattern Based on Grid Clique–
An algorithm is presented, which takes grid clique as a whole to find its all instance points neighbor lists in 9-neighbor area, and inserts them into the candidate extension clique list, then deletes the records that dont meet the criteria in the list to generate maximal clique.
- Research on Graphic Displaying Algorithm for Electron Beam Lithography–
The result shows that the displaying algorithm not only guarantees the precision of the graphic data, but also displays files with large amount of graphic data fast, improving the efficiency effectively.
- –
- –
- –
- Boundary Point Recognition Algorithm Based on Grid Adjacency Relation–
The experimental results show that the algorithm can recognize boundary points effectively and efficiently in various datasets with arbitrary shape and different size and density or with noise.
- Mesurer et améliorer la qualité des corpus comparables–
L'idee est avantageuse puisqu'elle peut etre utilisee avec n'importe quelle methode existante reposant sur des corpus comparables dans le but d'ameliorer les performances des applications qui les exploitent, nous developpons de nouveaux modeles CLIR en etendant les recents modeles proposes en recherche d'information monolingue.
- –
- Prototype of Virus Detection on Android Smartphone–
This paper presented framework for a light-weight anti-virus program for smartphone considering hardware limitations, a prototype program is developed on Android, and the efficiency test shows positive performance efficiency.
- –
- –
- Research of Decision-Making in the Multi-Agent System Based on Interactive Influence Diagrams–
This paper demonstrates that one can choose a minimal, but sufficient, space of mental models by combining models that action or utility equivalence applies, and provides empirical results by showing the minimal mental model spaces that it generates.
- Discovery of Protein’s Multifunction and Diversity of Information Transmission–
To clarify the manifold roles of a protein among different protein complexes and to reveal the diversity of information transmission among these complexes, essential elements, tasks, process and experiment are represented.
- –
- A New Vector Identification Method Based on Vector Structure and Suffix Tree–
The vector structures of ESTs are described and a new vector identification method combining vector structure and fast sequence alignment method based on suffix tree is presented, characterized by building a suffix tree through EST vector structure to carry out vector identification.
- A method of accelerating seismic Pre-stack time migration by GPU–
The General Purpose GPU technology is introduced and a method to realize pre-stack time migration software on GPU is proposed, which greatly improves computational efficiency, and then dramatically save power and maintenances cost.
- –
- –
- –
- Objective research and application of multi-channel human meridian impedance dynamic testing system–
This paper builds a multi-channel ring electrode system to point impedance testing of the human body for the first time, which bases on the hardware platform of AT89C52 combined with M82C54-2 and collects some real data of human body with this vehicle, whose results are expected to be more satisfactory in the near future.
- –
- Measuring Web Feature Impacts in BitTorrent-like Systems–
A systematically quantitative study on the impact of these attributes presented in the form of web features, by measuring the largest BitTorrent website in CERNET, confirms the above qualitative speculation and shows the implications of the web feature impact for system designers and potential attackers.
- The Digital Image Experiment Platform Based on Even Liner Grammar–
A digital image processing experiment platform which is on the theory basis of the Even Linear Grammar, based on the Automatization of COM, which can increase the reusability of the image processing methods.
- –
- A novel dynamic risk assessment for communication network system software–
Field data test results show that the DNRISK measure can provide a much better indicator for type II prediction error for interactive software modules and can be a reliable indicator to assist software engineers to effectively improve their products.
- –
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 2026-10-14. 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.