AAAI 2023 Papers — Page 13
AAAI Conference on Artificial Intelligence · 1578 papers
RETRACTED: McOmet: Multimodal Fusion Transformer for Physical Audiovisual Commonsense Reasoning
Daoming Zong (East China Normal University), Shiliang Sun (East China Normal University)
ClassificationRecognitionTransformerVideoTextMultimodalityAudio
🎯 What it does: A framework for multimodal audio-video joint reasoning of physical common sense (MCOMET) is proposed, and question answering and material classification tasks are implemented on the PACS dataset.
Retrosynthesis Prediction with Local Template Retrieval
Shufang Xie (Renmin University of China), Tao Qin (Microsoft Research)
RetrievalDrug DiscoveryGraph Neural NetworkGraph
🎯 What it does: This paper proposes a backward synthesis prediction method called RetroKNN based on KNN retrieval of local reaction templates, which combines GNN with retrieval results to generate the final prediction.
Reviewing Labels: Label Graph Network with Top-k Prediction Set for Relation Extraction
Bo Li (Peking University), Shikun Zhang (Peking University)
Graph Neural NetworkLarge Language ModelContrastive LearningText
🎯 What it does: The research utilizes a model to predict the Top-k label set to construct a label graph network, improving relation extraction; the KLG model is proposed;
Revisiting Classifier: Transferring Vision-Language Models for Video Recognition
Wenhao Wu (University of Sydney), Wanli Ouyang (University of Sydney)
ClassificationRecognitionTransformerVision Language ModelVideo
🎯 What it does: This paper proposes replacing the linear classifier in video classification with text embeddings from a pre-trained vision-language model, freezing the classifier weights, and only fine-tuning the visual encoder to enhance the model's transfer performance.
Revisiting Denoising Diffusion Probabilistic Models for Speech Enhancement: Condition Collapse, Efficiency and Refinement
Wenxin Tai (University of Electronic Science and Technology of China), Ting Zhong (University of Electronic Science and Technology of China)
RestorationGenerationDiffusion modelAudio
🎯 What it does: The DR-DiffuSE framework is proposed, which achieves high-quality speech enhancement through a conditional diffusion model, addressing issues such as conditional collapse, low efficiency, and degradation of audio quality after enhancement.
Revisiting the Spatial and Temporal Modeling for Few-Shot Action Recognition
Jiazheng Xing (Zhejiang University), Boyu Mu (Zhejiang University)
RecognitionConvolutional Neural NetworkVideo
🎯 What it does: Proposes the SloshNet framework, which integrates low-level and high-level spatial features, as well as long-term and short-term temporal modeling, to achieve few-shot action recognition.
Revisiting Unsupervised Local Descriptor Learning
Wufan Wang (Beijing Institute of Technology), Hua Huang (Beijing Normal University)
OptimizationRepresentation LearningHyperparameter SearchConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: A framework for unsupervised local descriptor learning named HybridDesc is proposed, which integrates two tuple construction methods: regularization and clustering, and enhances learning speed and performance through differentiable hyperparameter search (DHS) and on-demand clustering (ODC).
Reward Poisoning Attacks on Offline Multi-Agent Reinforcement Learning
Young Wu (University of Wisconsin Madison), Qiaomin Xie (University of Wisconsin Madison)
Adversarial AttackReinforcement LearningTabular
🎯 What it does: This paper studies reward pollution attacks in offline multi-agent reinforcement learning and designs a framework to minimize attack costs, allowing the attacker to modify the training set rewards so that each agent learns the target strategy specified by the attacker.
Reward-Based Negotiating Agent Strategies
Ryota Higa (NEC Corporation), Shinji Nakadai (NEC Corporation)
Reinforcement LearningAgentic AI
🎯 What it does: The research proposes a reward-based multi-issue negotiation agent strategy, implementing end-to-end reinforcement learning using an issue-based representation deep policy network, without the need for manually crafted utility functions.
Reward-Biased Maximum Likelihood Estimation for Neural Contextual Bandits: A Distributional Learning Perspective
Yu-Heng Hung (National Yang Ming Chiao Tung University), Ping-Chun Hsieh (National Yang Ming Chiao Tung University)
Recommendation SystemReinforcement LearningTabular
🎯 What it does: We propose NeuralRBMLE, a neural contextual bandit algorithm based on Reward Bias Maximum Likelihood Estimation (RBMLE), which utilizes neural networks to learn the likelihood of unknown reward distributions and directly explores in the parameter space. Two implementation schemes are provided: the gradient ascent version (NeuralRBMLE-GA) and the NTK-based approximate version (NeuralRBMLE-PC), and we prove its sublinear regret upper bound under the NTK approximation.
RGBD1K: A Large-Scale Dataset and Benchmark for RGB-D Object Tracking
Xue-Feng Zhu (Jiangnan University), Josef Kittler (University of Surrey)
Object TrackingTransformerImageVideoBenchmark
🎯 What it does: A large RGB-D tracking dataset called RGBD1K is proposed, and a baseline tracker named SPT is designed based on Transformer.
Rich Event Modeling for Script Event Prediction
Long Bai (Chinese Academy of Sciences), Xueqi Cheng (Chinese Academy of Sciences)
TransformerText
🎯 What it does: This paper proposes the Rich Event Prediction framework, which achieves script event prediction through rich event descriptions and Transformer encoding.
Riemannian Local Mechanism for SPD Neural Networks
Ziheng Chen (Jiangnan University), Josef Kittler (University of Surrey)
RecognitionVideo
🎯 What it does: A multi-scale submanifold network (MSNet) is designed and implemented, utilizing submatrix selection and logarithmic mapping to extract local geometric information in SPD networks.
RINK: Reader-Inherited Evidence Reranker for Table-and-Text Open Domain Question Answering
Eunhwan Park (Jeonbuk National University), Seung-Hoon Na (Naver Corporation)
RetrievalTransformerPrompt EngineeringTextTabular
🎯 What it does: This paper proposes an integrated reranker based on Reader (RINK) and embeds it into the Retriever-Reranker-Reader framework to enhance the performance of mixed retrieval and question answering involving tables and text.
RLEKF: An Optimizer for Deep Potential with Ab Initio Accuracy
Siyu Hu (Chinese Academy of Sciences), Tong Zhao (Chinese Academy of Sciences)
OptimizationTabularPhysics RelatedStochastic Differential Equation
🎯 What it does: A new optimizer RLEKF is proposed, specifically designed to accelerate the training of Deep Potential (DP) neural network potential energy surfaces, and experimental validation has been conducted on 13 different metal/semiconductor/insulator systems.
RLogist: Fast Observation Strategy on Whole-Slide Images with Deep Reinforcement Learning
Boxuan Zhao (Shanghai Jiao Tong University), Wei Yang (Tencent AI Lab)
ClassificationOptimizationConvolutional Neural NetworkTransformerReinforcement LearningImageBiomedical Data
🎯 What it does: A fast observation strategy RLogist based on deep reinforcement learning is proposed for quickly locating and classifying diagnosis-related areas on Whole-slide Images (WSI).
Robust and Fast Measure of Information via Low-Rank Representation
Yuxin Dong (Xi'an Jiaotong University), Chen Li (Xi'an Jiaotong University)
Computational EfficiencyTabular
🎯 What it does: A new information metric based on low-rank representation of R' enyi entropy is proposed, aimed at improving robustness to noise and computational efficiency.
Robust Causal Graph Representation Learning against Confounding Effects
Hang Gao (Institute of Software Chinese Academy of Sciences), Fuchun Sun (Tsinghua University)
Representation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Proposes RCGRL, which utilizes actively generated instrumental variables to achieve graph representation learning while removing confounding effects;
Robust Domain Adaptation for Machine Reading Comprehension
Liang Jiang (Ant Group), Xi Peng (Sichuan University)
Domain AdaptationTransformerReinforcement LearningText
🎯 What it does: A robust domain adaptation method RMRC is proposed, which constructs high-quality pseudo QA pairs using an answer extractor, question selector, and MRC model to address the noise correspondence problem in cross-domain reading comprehension.
Robust Feature Rectification of Pretrained Vision Models for Object Recognition
Shengchao Zhou (Institute of Automation Chinese Academy of Sciences), Shiming Xiang (Institute of Automation Chinese Academy of Sciences)
RecognitionRestorationObject DetectionConvolutional Neural NetworkImage
🎯 What it does: The ROFER module is proposed, which can combat unseen degradations (blurring, noise, low contrast) on pre-trained visual models and handle composite degradations.
Robust Image Denoising of No-Flash Images Guided by Consistent Flash Images
Geunwoo Oh (Gwangju Institute of Science and Technology), Bochang Moon (Gwangju Institute of Science and Technology)
RestorationConvolutional Neural NetworkImage
🎯 What it does: This paper proposes a robust image denoising framework that utilizes pairs of flash and no-flash images to achieve high-quality denoising under inconsistent lighting conditions.
Robust Multi-Agent Coordination via Evolutionary Generation of Auxiliary Adversarial Attackers
Lei Yuan (Nanjing University), Yang Yu (Nanjing University)
Adversarial AttackReinforcement LearningSequential
🎯 What it does: This paper proposes a robust cooperative multi-agent reinforcement learning framework called ROMANCE, specifically designed to address the policy disturbance problem under limited strategic opponents (limited attack attempts).
Robust Neuro-Symbolic Goal and Plan Recognition
Leonardo Amado (Pontifical Catholic University of Rio Grande do Sul), Felipe Meneguzzi (University of Aberdeen)
RecognitionOptimizationRecurrent Neural NetworkAuto EncoderSequential
🎯 What it does: A neural-symbolic framework for target and plan recognition, PPR, has been developed, utilizing predictive models to fill in missing or noisy observations while simultaneously inferring targets and complete plans.
Robust One-Shot Segmentation of Brain Tissues via Image-Aligned Style Transformation
Jinxin Lv (Huazhong University of Science and Technology), Qiang Li (Huazhong University of Science and Technology)
SegmentationConvolutional Neural NetworkImageBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: A single-sample brain tissue segmentation framework combining Image Alignment Style Transfer (IST) and Feature-aware Content Consistency (FCC) is proposed to address the issue of label and image space mismatch in traditional dual-model iterations.
Robust Representation Learning by Clustering with Bisimulation Metrics for Visual Reinforcement Learning with Distractions
Qiyuan Liu (University of Science and Technology of China), Jie Wang (University of Science and Technology of China)
Representation LearningReinforcement LearningImage
🎯 What it does: This paper proposes a clustering method based on a dual-track metric, CBM, for learning robust representations against dispersed distractions such as background, color, and camera perspective in visual reinforcement learning.
Robust Self-Supervised Multi-Instance Learning with Structure Awareness
Yejiang Wang (Northeastern University), Meixia Wang (Singapore Institute of Technology)
Representation LearningGraph Neural NetworkContrastive LearningTextGraph
🎯 What it does: This paper proposes a self-supervised multi-instance learning framework called SMILES, which utilizes data augmentation and contrastive learning to learn unlabeled bag representations and achieves structure awareness through learning trainable graph structures.
Robust Temporal Smoothness in Multi-Task Learning
Menghui Zhou (Yunnan University), Po Yang (Sheffield University)
Anomaly DetectionOptimizationTabularTime SeriesBiomedical DataAlzheimer's DiseaseAgriculture Related
🎯 What it does: The study introduces a Robust Time Smoothing framework (RoTS) in multi-task learning, which decomposes each task into smooth and discrete parts, enabling the identification and handling of anomalous tasks while maintaining temporal similarity.
Robust Video Portrait Reenactment via Personalized Representation Quantization
Kaisiyuan Wang (University of Sydney), Jingdong Wang (Baidu Inc.)
GenerationData SynthesisTransformerAuto EncoderVideo
🎯 What it does: A video avatar reenactment framework named VPNQ is proposed, which utilizes a personalized quantized local patch dictionary and spatiotemporal Transformer to achieve high-quality and robust video avatar generation under arbitrary driving videos.
RobustLoc: Robust Camera Pose Regression in Challenging Driving Environments
Sijie Wang (Continental-NTU Corporate Lab Nanyang Technological University), Diego Navarro Navarro (Continental Automotive Singapore)
Pose EstimationAutonomous DrivingConvolutional Neural NetworkGraph Neural NetworkImageVideoOrdinary Differential Equation
🎯 What it does: This paper proposes a robust camera pose regression model called RobustLoc, based on multi-view graph neural diffusion and neural differential equations, aimed at achieving high-precision localization in challenging driving environments affected by seasonal changes, weather, lighting, and dynamic object disturbances.
ROIFormer: Semantic-Aware Region of Interest Transformer for Efficient Self-Supervised Monocular Depth Estimation
Daitao Xing (New York University), Anthony Tzes (New York University)
Depth EstimationAutonomous DrivingTransformerImage
🎯 What it does: We propose ROIFormer, a semantic-guided local adaptive attention module for feature fusion and enhancement in self-supervised monocular depth estimation.
Rolling Horizon Based Temporal Decomposition for the Offline Pickup and Delivery Problem with Time Windows
Youngseo Kim (Cornell University), Samitha Samaranayake (Pennsylvania State University)
OptimizationTabular
🎯 What it does: A time decomposition framework based on rolling time slots is proposed for the offline Pickup and Delivery Problem with Time Windows (PDPTW). An online high-capacity passenger allocation algorithm is used to solve subproblems within each time slot, gradually stitching together to obtain the global vehicle route.
RPA: Reasoning Path Augmentation in Iterative Retrieving for Multi-Hop QA
Ziyi Cao (Harbin Institute of Technology), Shaobo Li (Harbin Institute of Technology)
RetrievalTransformerTextRetrieval-Augmented Generation
🎯 What it does: A multi-hop question answering retrieval method RPA based on reasoning path reordering and incremental generation is proposed.
RSPT: Reconstruct Surroundings and Predict Trajectory for Generalizable Active Object Tracking
Fangwei Zhong (Peking University), Yizhou Wang (Peking University)
Object TrackingRobotic IntelligenceRecurrent Neural NetworkReinforcement LearningImage
🎯 What it does: A RSPT framework is proposed to achieve generalized active target tracking by reconstructing the surrounding environment and predicting target trajectories.
Rule Induction in Knowledge Graphs Using Linear Programming
Sanjeeb Dash (IBM Research), Joao Goncalves (IBM Research)
OptimizationExplainability and InterpretabilityGraphBiomedical Data
🎯 What it does: A rule induction method based on linear programming is proposed to select and weight a compact and interpretable set of rules from knowledge graphs for knowledge graph completion.
Runtime Analysis for the NSGA-II: Provable Speed-Ups from Crossover
Benjamin Doerr (Ecole Polytechnique), Zhongdi Qu (Ecole Polytechnique)
OptimizationBenchmark
🎯 What it does: After introducing crossover operations to the multi-objective evolutionary algorithm NSGA-II, theoretical analysis and experimental validation demonstrated its speed improvement on the ONEJUMPZEROJUMP benchmark.
RWEN-TTS: Relation-Aware Word Encoding Network for Natural Text-to-Speech Synthesis
Shinhyeok Oh (Netmarble AI Center), Insoo Oh (Netmarble AI Center)
GenerationData SynthesisRecurrent Neural NetworkGraph Neural NetworkLarge Language ModelTextAudio
🎯 What it does: This paper proposes a Relation-Aware Word Encoding Network (RWEN) to improve the naturalness and expressiveness in text-to-speech (TTS) synthesis. The network consists of two parts: Semantic Relation Encoding (SRE) and Adjacent Word Relation Encoding (AWRE), which effectively utilize syntactic tree and contextual information at the word level.
Safe Interval Path Planning with Kinodynamic Constraints
Zain Alabedeen Ali (Moscow Institute of Physics and Technology), Konstantin Yakovlev (Federal Research Center for Computer Science and Control of Russian Academy of Sciences)
OptimizationRobotic IntelligenceSimultaneous Localization and MappingTabularBenchmark
🎯 What it does: A complete and optimal variant of SIPP (SIPP-IP) is proposed, capable of planning single-robot paths in dynamic obstacle environments while considering kinodynamic constraints.
Safe Multi-View Deep Classification
Wei Liu (Tongji University), Shaorong Xie (Shanghai University)
ClassificationSafty and PrivacyAuto EncoderImageVideo
🎯 What it does: A secure multi-view deep classification method (SMDC) is proposed, which ensures that the classification performance does not degrade when new views are added by fusing multiple views at the evidence level and minimizing uncertainty.
Safeguarded Learned Convex Optimization
Howard Heaton (Typal LLC), Wotao Yin (Alibaba US)
OptimizationImage
🎯 What it does: Proposes the Safe-L2O framework, which combines learning-driven algorithms (L2O) with safety protection in convex optimization to ensure convergence;
SAH: Shifting-Aware Asymmetric Hashing for Reverse k Maximum Inner Product Search
Qiang Huang (National University of Singapore), Anthony K. H. Tung (National University of Singapore)
Recommendation SystemRetrieval-Augmented Generation
🎯 What it does: A sub-quadratic time algorithm SAH (Shifting-Aware Hashing) is proposed for the reverse k maximum inner product search (RkMIPS) problem in high-dimensional data.
Scalable and Effective Conductance-Based Graph Clustering
Longlong Lin (Southwest University), Tao Jia (Southwest University)
Graph Neural NetworkGraph
🎯 What it does: A general framework PCon based on peeling is proposed, and two linear time/space graph clustering algorithms are designed under this framework: PCon core (using core-decomposition) and PCon de (using degree-based greedy peeling).
Scalable and Globally Optimal Generalized L₁ K-center Clustering via Constraint Generation in Mixed Integer Linear Programming
Aravinth Chembu (University of Toronto), Akshat Kumar (Singapore Management University)
Anomaly DetectionOptimizationTabular
🎯 What it does: This paper proposes a mixed-integer linear programming (MILP) model for the generalized L1 k-center clustering problem, achieving global optimal solutions for millions of samples through constraint generation techniques; it also incorporates outlier handling in the model.
Scalable Attributed-Graph Subspace Clustering
Chakib Fettal (Universite Paris Cite), Mohamed Nadif (Universite Paris Cite)
Computational EfficiencyGraph Neural NetworkGraph
🎯 What it does: This paper proposes a scalable attribute graph subspace clustering algorithm (SAGSC), which learns initial representations through graph convolutional networks, then constructs a non-negative similarity matrix using a factorized self-representation matrix and kernel mapping, and finally performs spectral clustering implicitly on this matrix.
Scalable Bayesian Meta-Learning through Generalized Implicit Gradients
Yilang Zhang (University of Minnesota), Georgios B. Giannakis (University of Minnesota)
OptimizationMeta LearningImage
🎯 What it does: This paper proposes a general implicit Bayesian meta-learning (iBaML) framework that efficiently computes the gradients of Bayesian meta-learning using implicit differentiation and the conjugate gradient method, avoiding the complexity bottleneck caused by explicit higher-order derivatives.
Scalable Decision-Focused Learning in Restless Multi-Armed Bandits with Application to Maternal and Child Health
Kai Wang (Harvard University), Milind Tambe (Google Research)
OptimizationReinforcement Learning from Human FeedbackReinforcement LearningTabularBiomedical DataElectronic Health Records
🎯 What it does: This paper proposes a decision-focused learning framework for restless multi-armed bandits (RMAB) with unknown transition dynamics, utilizing a differentiable Whittle index strategy to directly train predictive models to maximize policy performance.
Scalable Edge Blocking Algorithms for Defending Active Directory Style Attack Graphs
Mingyu Guo (University of Adelaide), Hung Nguyen (University of Adelaide)
OptimizationReinforcement LearningGraph
🎯 What it does: The study focuses on the Stackelberg game between attackers and defenders in the Active Directory attack graph, proposing an algorithm that can block a limited number of edges to minimize the attacker's success rate.
Scalable Optimal Multiway-Split Decision Trees with Constraints
Shivaram Subramanian (IBM Research), Wei Sun (IBM Research)
ClassificationOptimizationExplainability and InterpretabilityReinforcement LearningTabular
🎯 What it does: A path-based mixed integer programming framework based on column generation is proposed to learn multi-way split decision trees that can incorporate various constraints.
Scalable Spatial Memory for Scene Rendering and Navigation
Wen-Cheng Chen (National Cheng Kung University), Min-Chun Hu (National Tsing Hua University)
Convolutional Neural NetworkReinforcement LearningImage
🎯 What it does: This paper proposes the Scene Memory Network (SMN), which enables online scene representation and rendering without the need for depth maps or camera intrinsic parameters, and applies it to navigation and exploration in reinforcement learning.
Scalable Spatiotemporal Graph Neural Networks
Andrea Cini (Università della Svizzera italiana), Cesare Alippi (Politecnico di Milano)
Graph Neural NetworkGraphTime SeriesBenchmark
🎯 What it does: This paper proposes a scalable spatiotemporal graph neural network (SGP) that utilizes a Deep Echo State Network (random recurrent network) to perform multi-scale encoding of each sensor's time series. Information is then propagated and aggregated in the spatial dimension using a graph shift operator, ultimately resulting in spatiotemporal embeddings for the nodes. This embedding can be computed during the preprocessing stage, after which only a lightweight multi-layer perceptron decoder needs to be trained to complete the predictions.
Scalable Theory-Driven Regularization of Scene Graph Generation Models
Davide Buffelli (University of Padova), Efthymia Tsamoura (Samsung AI)
GenerationOptimizationGraph Neural NetworkGraph
🎯 What it does: A scalable neural symbolic regularization framework NGPU is proposed, which constrains the scene graph generation model during training using negative consistency constraints, enhancing the model's consistency and generalization ability regarding knowledge.
Scaling Law for Recommendation Models: Towards General-Purpose User Representations
Kyuyong Shin (NAVER), Kyung-Min Kim (NAVER)
Recommendation SystemTransformerContrastive LearningMultimodality
🎯 What it does: A general user representation model CLUE is proposed and trained, which learns universal user embeddings from large-scale user behavior data across services using contrastive learning.
Scaling Marginalized Importance Sampling to High-Dimensional State-Spaces via State Abstraction
Brahma S. Pavse (University of Wisconsin - Madison), Josiah P. Hanna (University of Wisconsin - Madison)
Reinforcement Learning
🎯 What it does: This paper improves the offline evaluation performance of marginal importance sampling (MIS) in offline reinforcement learning by mapping high-dimensional state spaces to low-dimensional abstract state spaces.
Scaling Up Dynamic Graph Representation Learning via Spiking Neural Networks
Jintang Li (Sun Yat-sen University), Changhua Meng (Ant Group)
Representation LearningRecurrent Neural NetworkGraph Neural NetworkSpiking Neural NetworkGraph
🎯 What it does: Proposes the SpikeNet framework, which uses SNN to efficiently learn large-scale time-varying graph node representations instead of RNN.
ScatterFormer: Locally-Invariant Scattering Transformer for Patient-Independent Multispectral Detection of Epileptiform Discharges
Ruizhe Zheng (Fudan University), Yuguo Yu (Fudan University)
ClassificationAnomaly DetectionTransformerBiomedical Data
🎯 What it does: A hierarchical Transformer ScatterFormer based on invariant scattering transform is proposed for patient-independent epilepsy seizure detection.
Scene Graph to Image Synthesis via Knowledge Consensus
Yang Wu (Sun Yat-sen University), Liang Lin (Sun Yat-sen University)
GenerationData SynthesisGraph Neural NetworkAuto EncoderGenerative Adversarial NetworkImage
🎯 What it does: This paper proposes a method for image generation that solely utilizes scene graphs, avoiding the use of auxiliary information such as layouts or segmentations.
Scene-Level Sketch-Based Image Retrieval with Minimal Pairwise Supervision
Ce Ge (Beijing University of Posts and Telecommunications), Jianxin Liao (Beijing University of Posts and Telecommunications)
RetrievalGraph Neural NetworkAuto EncoderImage
🎯 What it does: This paper studies the scene-level sketch retrieval task, using only pairwise supervision for retrieval learning.
SCI: A Spectrum Concentrated Implicit Neural Compression for Biomedical Data
Runzhao Yang (Tsinghua University), Qionghai Dai (Tsinghua University)
CompressionOptimizationBiomedical DataComputed Tomography
🎯 What it does: Proposed the SCI method for implicit neural compression of diverse biomedical data.
Score-Based Learning of Graphical Event Models with Background Knowledge Augmentation
Debarun Bhattacharjya (IBM), Xiao Shou (Rensselaer Polytechnic Institute)
Graph Neural NetworkScore-based ModelGraphBiomedical Data
🎯 What it does: This study investigates a scheme to enhance the learning of graphical event models (GEM) in low data environments through background knowledge.
Script, Language, and Labels: Overcoming Three Discrepancies for Low-Resource Language Specialization
Jaeseong Lee (Seoul National University), Seung-won Hwang (Seoul National University)
TransformerSupervised Fine-TuningText
🎯 What it does: This paper proposes a solution to three types of mismatches (script, language, label) encountered when specializing multilingual pre-trained models for low-resource languages, along with corresponding solutions.
SEAT: Stable and Explainable Attention
Lijie Hu (King Abdullah University of Science and Technology), Di Wang (King Abdullah University of Science and Technology)
OptimizationExplainability and InterpretabilityRecurrent Neural NetworkTransformerText
🎯 What it does: Design and propose Stable Explainable Attention (SEAT) by defining three main attributes: stability, explanation overlap, and prediction similarity, and based on this, construct a min-max optimization framework to obtain a more robust attention distribution.
Second-Order Quantified Boolean Logic
Jie-Hong R. Jiang (National Taiwan University)
🎯 What it does: This paper studies second-order quantified Boolean logic (SOQBF) and proposes a quantifier elimination method to convert SOQBF into an equivalent QBF, a game-theoretic semantic interpretation of SOQBF, a complete proof system SOQ-res, and an algorithm for extracting Herbrand functions from the proof. It also discusses the potential applications of SOQBF in system design and multi-agent planning.
Securing Lifelines: Safe Delivery of Critical Services in Areas with Volatile Security Situation via a Stackelberg Game Approach
Tien Mai (Singapore Management University), Arunesh Sinha (Rutgers University)
OptimizationSafty and PrivacyReinforcement Learning
🎯 What it does: A Stackelberg security game model is proposed to address the placement of limited vaccination centers in areas with high security risks, while considering both center selection and security resource allocation.
Securing Secure Aggregation: Mitigating Multi-Round Privacy Leakage in Federated Learning
Jinhyun So (University of Southern California), A. Salman Avestimehr (University of Southern California)
Federated LearningSafty and PrivacyConvolutional Neural NetworkImage
🎯 What it does: A multi-round secure aggregation framework named Multi-RoundSecAgg is proposed to address the long-term privacy leakage issues caused by partial user participation in cross-device federated learning.
SeDepTTS: Enhancing the Naturalness via Semantic Dependency and Local Convolution for Text-to-Speech Synthesis
Chenglong Jiang (South China University of Technology), Hongzhong Zhen (South China University of Technology)
GenerationData SynthesisTransformerAudio
🎯 What it does: This paper proposes SeDepTTS, a non-autoregressive text-to-speech model based on FastSpeech2, which integrates semantic dependency priors and local convolution to enhance local modeling of attention;
See How You Read? Multi-Reading Habits Fusion Reasoning for Multi-Modal Fake News Detection
Lianwei Wu (Northwestern Polytechnical University), Yanning Zhang (Northwestern Polytechnical University)
ClassificationAnomaly DetectionConvolutional Neural NetworkRecurrent Neural NetworkLarge Language ModelTextMultimodality
🎯 What it does: Proposes a Multi-Reading Habit Fusion Reasoning Network (MRHFR) for multimodal fake news detection.
See Your Emotion from Gait Using Unlabeled Skeleton Data
Haifeng Lu (Lanzhou University), Bin Hu (Beijing Institute of Technology)
RecognitionGraph Neural NetworkContrastive LearningTime Series
🎯 What it does: This paper proposes a self-supervised gait emotion recognition framework called CAGE, which learns gait expression features using cross-coordinate contrastive learning and fuzzy sample augmentation based on skeletal sequences.
SEFormer: Structure Embedding Transformer for 3D Object Detection
Xiaoyu Feng (Tsinghua University), Yongpan Liu (National University of Singapore)
Object DetectionAutonomous DrivingTransformerPoint Cloud
🎯 What it does: This paper proposes a Structure-Embedded Transformer (SEFormer) and constructs a multi-scale SEFormer network for LiDAR 3D object detection, which can retain and encode local structural information in sparse point clouds.
SegFormer: A Topic Segmentation Model with Controllable Range of Attention
Haitao Bai (Xi'an Jiaotong University), Zhou Su (Xi'an Jiaotong University)
ClassificationSegmentationTransformerText
🎯 What it does: This paper presents SegFormer, an end-to-end neural topic segmentation model that utilizes unidirectional attention blocks, a context aggregator, and topic classification loss.
SelectAugment: Hierarchical Deterministic Sample Selection for Data Augmentation
Shiqi Lin (University of Science and Technology of China), Zhibo Chen (University of Science and Technology of China)
Domain AdaptationData-Centric LearningReinforcement LearningImage
🎯 What it does: This study investigates the distribution shift problem caused by random sample selection during data augmentation and proposes a deterministic sample selection method called SelectAugment based on hierarchical reinforcement learning.
Selective Knowledge Distillation for Non-Autoregressive Neural Machine Translation
Min Liu (Nanjing University), Shujian Huang (ByteDance AI Lab)
Knowledge DistillationTransformerText
🎯 What it does: A selective knowledge distillation method is proposed, utilizing a NAT evaluator to dynamically replace target sentences, reducing teacher error while combining with the original translation.
Selector-Enhancer: Learning Dynamic Selection of Local and Non-local Attention Operation for Speech Enhancement
Xinmeng Xu (Wuhan University), Yuhong Yang (Wuhan University)
RestorationConvolutional Neural NetworkRecurrent Neural NetworkReinforcement LearningAudio
🎯 What it does: Proposes Selector-Enhancer, which utilizes dual attention CNN and feature filters to dynamically select local and global attention for speech enhancement.
Self Correspondence Distillation for End-to-End Weakly-Supervised Semantic Segmentation
Rongtao Xu (Institute of Automation, Chinese Academy of Sciences), Xiaopeng Zhang (Institute of Automation, Chinese Academy of Sciences)
SegmentationKnowledge DistillationTransformerImage
🎯 What it does: This paper proposes an end-to-end weakly supervised semantic segmentation framework called TSCD, which is centered around Self Correspondence Distillation (SCD) and a Variation-Aware Refine Module (VARM) based on pixel variation, aimed at refining CAM-based pseudo labels.
Self-Asymmetric Invertible Network for Compression-Aware Image Rescaling
Jinhai Yang (Bytedance Inc), Li Zhang (Bytedance Inc)
RestorationSuper ResolutionCompressionFlow-based ModelImage
🎯 What it does: A self-symmetric invertible network (SAIN) is proposed for compressed sensing image rescaling, capable of performing downsampling and simulating compression in a single forward pass, while recovering details in the reverse process.
Self-Contrastive Learning: Single-Viewed Supervised Contrastive Framework Using Sub-network
Sangmin Bae (Korea Advanced Institute of Science and Technology), Se-Young Yun (Korea Advanced Institute of Science and Technology)
ClassificationObject DetectionSegmentationRepresentation LearningConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: A Self-Contrastive (SelfCon) learning framework is proposed, which generates multiple features through a multi-exit network under a single view and compares them, achieving supervised contrastive learning without data augmentation.
Self-Decoupling and Ensemble Distillation for Efficient Segmentation
Yuang Liu (East China Normal University), Jun Wang (East China Normal University)
SegmentationComputational EfficiencyKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: This paper proposes the SDES framework, which enhances the performance of lightweight semantic segmentation networks through self-decoupling and integrated distillation.
Self-Emphasizing Network for Continuous Sign Language Recognition
Lianyu Hu (Tianjin University), Wei Feng (Tianjin University)
RecognitionConvolutional Neural NetworkRecurrent Neural NetworkVideo
🎯 What it does: A self-emphasizing network (SEN) is proposed, which highlights hand and facial information through lightweight spatial and temporal self-emphasizing modules to enhance continuous sign language recognition performance.
Self-Organization Preserved Graph Structure Learning with Principle of Relevant Information
Qingyun Sun (Beihang University), Philip S. Yu (University of Illinois at Chicago)
ClassificationRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: A graph structure learning framework called PRI-GSL based on the principle of relevant information is proposed, utilizing quantum continuous walks and multi-scale graph wavelet encoding of node roles to achieve self-organizing structure through unsupervised learning.
Self-Supervised Action Representation Learning from Partial Spatio-Temporal Skeleton Sequences
Yujie Zhou (Renmin University of China), Jiaqi Wang (Shanghai AI Laboratory)
Pose EstimationRepresentation LearningGraph Neural NetworkContrastive LearningVideo
🎯 What it does: This paper proposes a self-supervised skeletal sequence representation learning framework based on a three-stream structure (PSTL), which generates local masked samples through Central Space Masking (CSM) and Motion Attention Temporal Masking (MATM), and uses a cross-correlation matrix to learn the spatial-temporal relationships of the skeleton.
Self-Supervised Audio-Visual Representation Learning with Relaxed Cross-Modal Synchronicity
Pritam Sarkar (Queen's University), Ali Etemad (Queen's University)
ClassificationRecognitionRetrievalRepresentation LearningConvolutional Neural NetworkContrastive LearningVideoMultimodalityAudio
🎯 What it does: Proposes the CrissCross framework, which learns cross-modal representations of audio and video in a self-supervised manner, and relaxes the temporal synchronization requirements between audio and video by introducing asynchronous cross-modal loss.
Self-Supervised Bidirectional Learning for Graph Matching
Wenqi Guo (Shanghai Jiao Tong University), Lei Xu (Shanghai Jiao Tong University)
SegmentationOptimizationRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: A self-supervised bidirectional learning method IA-SSGM is proposed, which utilizes the Hungarian solver to generate pseudo-labels and enhances graph matching performance through cyclic consistency constraints.
Self-Supervised Continual Graph Learning in Adaptive Riemannian Spaces
Li Sun (North China Electric Power University), Philip S. Yu (University of Illinois at Chicago)
Knowledge DistillationRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Developed a self-supervised continuous graph learning framework RieGrace, capable of continuously learning new graph tasks in Riemannian spaces with varying curvatures.
Self-Supervised Graph Attention Networks for Deep Weighted Multi-View Clustering
Zongmo Huang (University of Electronic Science and Technology of China), Lifang He (Lehigh University)
Representation LearningGraph Neural NetworkAuto EncoderGraph
🎯 What it does: A deep weighted multi-view clustering framework based on self-supervised graph attention networks (SGDMC) is proposed, which constructs a k-NN graph and integrates self-supervised pseudo-labels in the graph attention layer to achieve a unified global and local assessment of node similarity.
Self-Supervised Graph Learning for Long-Tailed Cognitive Diagnosis
Shanshan Wang (Anhui University), Xingyi Zhang (University of Science and Technology of China)
Graph Neural NetworkContrastive LearningGraph
🎯 What it does: In the cognitive diagnosis task, a self-supervised graph learning framework SCD is introduced, which generates sparse views through importance-driven edge dropout on the original student-exercise interaction graph, and utilizes graph contrastive learning to assist the main task, enhancing the diagnostic effectiveness for long-tail students.
Self-Supervised Image Denoising Using Implicit Deep Denoiser Prior
Huangxing Lin (Xiamen University), John Paisley (Columbia University)
RestorationConvolutional Neural NetworkImage
🎯 What it does: This paper proposes a self-supervised image denoising method—Implicit Deep Denoiser Prior (IDDP), which utilizes the network's own output as a prior and achieves training without clean images through re-noising and regularization.
Self-Supervised Image Local Forgery Detection by JPEG Compression Trace
Xiuli Bi (Chongqing University of Posts and Telecommunications), Xinbo Gao (Chongqing University of Posts and Telecommunications)
CompressionAnomaly DetectionConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: This paper proposes a self-supervised image local forgery detection method based on JPEG compression traces, which includes a compression trace extractor and a contrastive learning module, capable of achieving pixel-level localization of various local forgeries without the need for a large amount of labeled data.
Self-Supervised Interest Transfer Network via Prototypical Contrastive Learning for Recommendation
Guoqiang Sun (Zhejiang University), Fei Fang (Alibaba Group)
Recommendation SystemContrastive LearningTabular
🎯 What it does: A self-supervised interest transfer network (SITN) is proposed for cross-domain recommendation tasks, achieving the transfer of interest knowledge across different domains through instance and cluster-level contrastive learning.
Self-Supervised Joint Dynamic Scene Reconstruction and Optical Flow Estimation for Spiking Camera
Shiyan Chen (Peking University), Tiejun Huang (Peking University)
RestorationSpiking Neural NetworkOptical FlowImageVideo
🎯 What it does: A self-supervised joint framework is proposed that can simultaneously reconstruct high-quality dynamic scene images and estimate optical flow from the asynchronous spike streams of a neuromorphic camera.
Self-Supervised Learning for Anomalous Channel Detection in EEG Graphs: Application to Seizure Analysis
Thi Kieu Khanh Ho (McGill University), Narges Armanfard (McGill University)
Anomaly DetectionGraph Neural NetworkAuto EncoderContrastive LearningGraphBiomedical Data
🎯 What it does: Construct EEG graphs and use self-supervised contrastive learning and generative learning to detect abnormal channels and segments during unannotated epileptic seizures.
Self-Supervised Learning for Multilevel Skeleton-Based Forgery Detection via Temporal-Causal Consistency of Actions
Liang Hu (Tongji University), Zhong Yuan Lai (University of Sydney)
Pose EstimationAnomaly DetectionGraph Neural NetworkContrastive LearningVideo
🎯 What it does: This paper proposes a self-supervised multi-layer skeletal forgery detection framework TC-SFDN, which can detect temporal inconsistencies and tampering in skeletal trajectories at the frame, segment, and action levels.
Self-Supervised Logic Induction for Explainable Fuzzy Temporal Commonsense Reasoning
Bibo Cai (Harbin Institute of Technology), Lifeng Shang (Huawei Noah's Ark Lab)
Explainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningTextSequential
🎯 What it does: A self-supervised logical induction framework called LECTER is proposed for fuzzy temporal commonsense reasoning, which can explicitly derive temporal dependencies and perform interpretability validation.
Self-Supervised Primal-Dual Learning for Constrained Optimization
Seonho Park (Georgia Institute of Technology), Pascal Van Hentenryck (Georgia Institute of Technology)
Optimization
🎯 What it does: A self-supervised dual learning (Primal-Dual Learning, PDL) framework is proposed, which directly approximates the optimal solution of constrained optimization problems using neural networks, without the need to pre-generate or solve instances.
Self-Supervised Video Representation Learning via Latent Time Navigation
Di Yang (Inria), François Brémond (Inria)
Representation LearningConvolutional Neural NetworkContrastive LearningVideo
🎯 What it does: A self-supervised video representation learning framework called Latent Time Navigation (LTN) is proposed, which performs contrastive learning on video segments by separating the time encoding subspace in the latent space and combining it with time parameterization.
Semantic 3D-Aware Portrait Synthesis and Manipulation Based on Compositional Neural Radiance Field
Tianxiang Ma (University of Chinese Academy of Sciences), Tieniu Tan (Chinese Academy of Sciences)
GenerationData SynthesisNeural Radiance FieldGenerative Adversarial NetworkImage
🎯 What it does: This paper proposes a neural radiance field for segmented semantic regions (CNeRF), achieving the synthesis of 3D consistent portraits and semantic-level editing.
Semantic-Aware Superpixel for Weakly Supervised Semantic Segmentation
Sangtae Kim (Seoul National University), Byonghyo Shim (Seoul National University)
SegmentationTransformerContrastive LearningImage
🎯 What it does: A method for semantic-aware superpixel discovery based on self-supervised visual Transformer (DINO) features is proposed, and this superpixel is used for seed expansion-based weakly supervised semantic segmentation training.
Semantic-Enhanced Image Clustering
Shaotian Cai (Shenzhen University), Longteng Chen (Shenzhen University)
ClassificationRepresentation LearningVision Language ModelContrastive LearningImage
🎯 What it does: This paper proposes an unsupervised image clustering method called SIC based on the CLIP visual-language pre-training model, which can map images to a semantic space and generate pseudo-labels without using category labels. It then jointly optimizes the clustering results through consistency learning in both image space and semantic space.
Semantics-Aware Dynamic Localization and Refinement for Referring Image Segmentation
Zhao Yang (Tsinghua University), Philip H.S. Torr (University of Oxford)
Object DetectionSegmentationTransformerVision Language ModelImage
🎯 What it does: A semantic-aware dynamic localization and refinement (SADLR) framework is proposed, which utilizes dynamic convolution to iteratively improve cross-modal features for reference image segmentation.
Semi-attention Partition for Occluded Person Re-identification
Mengxi Jia (Peking University), Ying Li (Peking University)
RecognitionSegmentationRetrievalKnowledge DistillationTransformerImage
🎯 What it does: A Semi-Attention Partition (SAP) framework is proposed, utilizing a noise semantic segmentation teacher to guide the attention learning of a Transformer student through knowledge distillation, thereby achieving more accurate body part segmentation and portrait features.
Semi-random Impossibilities of Condorcet Criterion
Lirong Xia (Rensselaer Polytechnic Institute)
🎯 What it does: This paper studies the impossibility problem of the Condorcet criterion (CC) in relation to four types of axioms: participation (PAR), half-monotonicity (HM), Maskin monotonicity (MM), and strategy-proofness (SP) under a semi-random model. By constructing and amplifying the classic worst-case proof diagram and combining it with concentration techniques, it is proven that under any distribution set Π that satisfies certain positivity and closure properties, for any voting rule with n voters and a group size B ≤ √n, there exists at least Ω(B√n) probability of a conflict between CC and any of the aforementioned axioms.
Semi-supervised Deep Large-Baseline Homography Estimation with Progressive Equivalence Constraint
Hai Jiang (Sichuan University), Shuaicheng Liu (University of Electronic Science and Technology of China)
Image TranslationOptimizationConvolutional Neural NetworkOptical FlowImage
🎯 What it does: An advanced semi-supervised homography estimation method is proposed, capable of achieving accurate homography estimation in large baseline scenarios.
Semi-Supervised Deep Regression with Uncertainty Consistency and Variational Model Ensembling via Bayesian Neural Networks
Weihang Dai (Hong Kong University of Science and Technology), Kwang-Ting Cheng (Hong Kong University of Science and Technology)
ImageUltrasound
🎯 What it does: This paper proposes a semi-supervised deep regression framework UCVME based on Bayesian neural networks, which utilizes uncertainty consistency loss and variational model ensemble to generate high-quality pseudo-labels, enhancing the performance of regression tasks.