AAAI 2023 Papers — Page 2
AAAI Conference on Artificial Intelligence · 1578 papers
Astromorphic Self-Repair of Neuromorphic Hardware Systems
Zhuangyu Han (Penn State University), Abhronil Sengupta (Penn State University)
ClassificationRecognitionSpiking Neural NetworkImage
🎯 What it does: This study investigates the self-repair mechanism of astrocytes in neuromorphic hardware and proposes a local astrocyte-modulated STDP learning rule based on a macro model.
Asynchronous Event Processing with Local-Shift Graph Convolutional Network
Linhui Sun (Institute of Automation, Chinese Academy of Sciences), Hanqing Lu (Institute of Automation, Chinese Academy of Sciences)
Anomaly DetectionAutonomous DrivingComputational EfficiencyGraph Neural NetworkGraph
🎯 What it does: The paper proposes a Local Shift Graph Convolutional Network (LSNet) for efficiently processing asynchronous sparse event streams from event cameras, and designs a node importance parallel pooling and asynchronous event processing mechanism.
Attack Can Benefit: An Adversarial Approach to Recognizing Facial Expressions under Noisy Annotations
Jiawen Zheng (Xiamen University), Shouhong Ding (Tencent)
RecognitionAdversarial AttackData-Centric LearningConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: This paper proposes a noise label detection and correction framework based on adversarial attacks, called GAAVE, to address the issues of noisy annotations and class imbalance in real-world FER datasets.
Attention-Based Depth Distillation with 3D-Aware Positional Encoding for Monocular 3D Object Detection
Zizhang Wu (ZongmuTech), Xiaoquan Wang (Fudan University)
Object DetectionDepth EstimationAutonomous DrivingKnowledge DistillationTransformerPoint CloudBenchmark
🎯 What it does: A deep distillation framework based on attention (ADD) is proposed, utilizing 3D-aware positional encoding for distillation in monocular 3D object detection to enhance 3D localization accuracy.
Attribute and Structure Preserving Graph Contrastive Learning
Jialu Chen (Southwestern University of Finance and Economics), Gang Kou (Southwestern University of Finance and Economics)
OptimizationRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: A new graph contrastive learning framework, ASP, is proposed, which utilizes a combination of the original view, attribute view, and high-order global structure view to achieve comprehensive retention of node attributes and structural information.
AUC Maximization for Low-Resource Named Entity Recognition
Ngoc Dang Nguyen (Monash University), Changyou Chen
RecognitionOptimizationTransformerSupervised Fine-TuningTextBiomedical Data
🎯 What it does: In response to the low-resource and class-imbalanced named entity recognition task, the authors transformed the traditional BIO annotation into two tasks (whether it belongs to an entity and whether it is the first word of the entity) and used an AUC maximization loss function for training.
Audio-Visual Contrastive Learning with Temporal Self-Supervision
Simon Jenni (Adobe Research), John Collomosse (University of Surrey)
ClassificationRecognitionRetrievalRepresentation LearningConvolutional Neural NetworkContrastive LearningVideoMultimodalityAudio
🎯 What it does: This paper proposes an unsupervised video and audio self-supervised learning framework that combines temporal self-supervision and cross-modal contrastive learning to learn a unified representation of video frames and corresponding audio.
AudioEar: Single-View Ear Reconstruction for Personalized Spatial Audio
Xiaoyang Huang (Shanghai Jiao Tong University), Teng Li (Shanghai Jiao Tong University)
GenerationDepth EstimationConvolutional Neural NetworkSupervised Fine-TuningImagePoint CloudMesh
🎯 What it does: Reconstructing a 3D ear model from a single view image, constructing a high-quality ear dataset (AudioEar3D and AudioEar2D), and combining the reconstructed ear with a 3D human model to simulate personalized HRTF for accurate spatial audio rendering.
Augmented Proximal Policy Optimization for Safe Reinforcement Learning
Juntao Dai (Zhejiang University), Gang Pan (Zhejiang University)
OptimizationSafty and PrivacyReinforcement Learning
🎯 What it does: This paper proposes a new safe reinforcement learning algorithm—Augmented Proximal Policy Optimization (APPO), which achieves precise control of constrained costs by adding a quadratic penalty term to the Lagrangian function;
Augmenting Affective Dependency Graph via Iterative Incongruity Graph Learning for Sarcasm Detection
Xiaobao Wang (Tianjin University), Jianwu Dang
ClassificationRecurrent Neural NetworkGraph Neural NetworkText
🎯 What it does: An Iterative Augmented Sentiment and Dependency Graph (IAAD) framework is proposed to improve sarcasm detection. This framework first constructs sentiment and dependency graphs based on SenticNet, and then iteratively learns through a parameterized semantic inconsistency measure and graph convolutional networks to dynamically update the graph structure and filter noise.
Auto-Weighted Multi-View Clustering for Large-Scale Data
Xinhang Wan (National University of Defense Technology), Lu Zhou (Nanjing University of Aeronautics and Astronautics)
OptimizationComputational EfficiencyImage
🎯 What it does: A self-weighted multi-view clustering method AWMVC is proposed, which utilizes multi-dimensional matrix decomposition and automatically weighted generation of consensus clustering matrices, suitable for large-scale data.
AutoGraph: Optimizing DNN Computation Graph for Parallel GPU Kernel Execution
Yuxuan Zhao (Chinese University of Hong Kong), Bei Yu (Chinese University of Hong Kong)
OptimizationComputational EfficiencyConvolutional Neural NetworkRecurrent Neural NetworkTransformerImage
🎯 What it does: This paper proposes the AutoGraph framework, which optimizes the computation graph of deep networks through dynamic programming + backtracking search and a hybrid critical path cost estimation, achieving efficient inference for GPU multi-stream parallel kernel scheduling.
AutoInit: Analytic Signal-Preserving Weight Initialization for Neural Networks
Garrett Bingham (University of Texas at Austin), Risto Miikkulainen (University of Texas at Austin)
OptimizationHyperparameter SearchConvolutional Neural NetworkTransformerImageTabular
🎯 What it does: This paper proposes an automated weight initialization algorithm called AutoInit, which can maintain the network's mean at zero and variance at one by analyzing the propagation of signal mean and variance under different network structures, activation functions, layer types, and hyperparameter settings, thus avoiding gradient explosion or vanishing.
Automata Cascades: Expressivity and Sample Complexity
Alessandro Ronca (Sapienza University of Rome), Giuseppe De Giacomo (University of Oxford)
🎯 What it does: A structured representation method for decomposing automata into serializable components (automata cascades) is proposed, and its expressive power and sample complexity are analyzed within this framework.
Automated Verification of Propositional Agent Abstraction for Classical Planning via CTLK Model Checking
Kailun Luo (Dongguan University of Technology)
Agentic AIBenchmark
🎯 What it does: The research addresses the existence problem of propositional agent abstraction given a refined mapping and provides an automatic verification method based on CTLK model checking.
Automated Verification of Social Laws in Numeric Settings
Ronen Nir (Technion Israel Institute of Technology), Erez Karpas (Technion Israel Institute of Technology)
OptimizationTabularBenchmark
🎯 What it does: The study verifies the robustness of the social method in a numerical planning environment, proposing a scheme to compile robustness verification into a single-agent numerical planning problem, and implements and evaluates it.
Automatically Verifying Expressive Epistemic Properties of Programs
Francesco Belardinelli (Imperial College London), Fortunat Rajaona (Surrey Centre for Cyber Security)
🎯 What it does: A new program-cognitive logic LPK is proposed, and a recursive translation from this logic to first-order logic is provided. Subsequently, this translation is implemented in Haskell and model checking of the program's knowledge properties is performed using an SMT solver.
AutoNF: Automated Architecture Optimization of Normalizing Flows with Unconstrained Continuous Relaxation Admitting Optimal Discrete Solution
Yu Wang (University of California), Peng Li (Oak Ridge National Laboratory)
OptimizationNeural Architecture SearchFlow-based ModelTabular
🎯 What it does: AutoNF automatically optimizes the hierarchical structure of Normalizing Flows through differentiable architecture search, maintaining reversibility while achieving a balance between computational efficiency and expressiveness.
AutoStegaFont: Synthesizing Vector Fonts for Hiding Information in Documents
Xi Yang (University of Science and Technology of China), Nenghai Yu (University of Science and Technology of China)
GenerationData SynthesisSafty and PrivacyAuto EncoderGenerative Adversarial NetworkImageText
🎯 What it does: This paper proposes AutoStegaFont, a dual-stage, dual-modal framework that can automatically generate vector fonts for document information hiding, while ensuring automation, generalization, and robustness.
AutoSTL: Automated Spatio-Temporal Multi-Task Learning
Zijian Zhang (Jilin University), Junbo Zhang (JD Technology)
Recurrent Neural NetworkGraph Neural NetworkTransformerTime Series
🎯 What it does: This paper proposes AutoSTL, an end-to-end automated spatiotemporal multi-task learning framework that can simultaneously predict various traffic attributes.
AVCAffe: A Large Scale Audio-Visual Dataset of Cognitive Load and Affect for Remote Work
Pritam Sarkar (Queen's University), Ali Etemad (Queen's University)
ClassificationRecognitionConvolutional Neural NetworkVideoMultimodalityBenchmarkAudio
🎯 What it does: The AVCAffe dataset has been constructed and made publicly available, recording 108 hours of audio-video from 106 subjects in a remote work context, divided into 58k segments by task, with self-reported emotional (arousal, valence) and cognitive load (mental demand, time demand, effort, etc.) labels.
Avocodo: Generative Adversarial Network for Artifact-Free Vocoder
Taejun Bak (NCSOFT), Young-Sun Joo (NCSOFT)
GenerationData SynthesisGenerative Adversarial NetworkAudio
🎯 What it does: A GAN-based speech synthesizer named Avocodo is proposed, aimed at eliminating imaging and aliasing noise caused by upsampling and downsampling in traditional GAN vocoders;
Back to the Future: Toward a Hybrid Architecture for Ad Hoc Teamwork
Hasra Dodampegama (University of Birmingham), Mohan Sridharan (University of Birmingham)
Explainability and InterpretabilityRobotic IntelligenceGraph Neural NetworkReinforcement LearningAgentic AISequential
🎯 What it does: A hybrid knowledge-based and data-driven architecture has been designed and implemented to achieve ad hoc teamwork in an uncoordinated multi-agent environment (Fort Attack); this architecture combines non-monotonic logical reasoning with a fast learning behavior prediction model, enabling quick adaptation and providing interpretable decisions when team members or tasks change.
Background-Mixed Augmentation for Weakly Supervised Change Detection
Rui Huang (Civil Aviation University of China), Yang Liu (Nanyang Technological University)
Data SynthesisConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: A background mixing-based data augmentation method (BGMix) and an incremental consistency loss with real data are proposed to construct a weakly supervised change detection framework that only requires image-level labels.
Backpropagation-Free Deep Learning with Recursive Local Representation Alignment
Alexander G. Ororbia (Rochester Institute of Technology), C. Lee Giles (Pennsylvania State University)
ClassificationOptimizationConvolutional Neural NetworkImage
🎯 What it does: A deep learning algorithm without backpropagation is proposed - Recursive Local Representation Alignment (rec-LRA) to train deep neural networks.
Balanced Column-Wise Block Pruning for Maximizing GPU Parallelism
Cheonjun Park (Yonsei University), Won Woo Ro (Ewha Womans University)
OptimizationComputational EfficiencyConvolutional Neural NetworkImageText
🎯 What it does: A pruning method based on GPU Block Column Pruning (BCBP) is proposed, utilizing the tiling structure of convolutional layers to prune by column blocks while maintaining the same sparsity rate for each block, thereby maximizing the utilization of GPU computing resources.
Balanced Meta Learning and Diverse Sampling for Lifelong Task-Oriented Dialogue Systems
Qiancheng Xu (Georgia Institute of Technology), Ruifeng Xu (Harbin Institute of Technology)
Meta LearningTransformerLarge Language ModelText
🎯 What it does: A two-stage lifelong task-oriented dialogue system MetaLTDS is proposed, which balances catastrophic forgetting suppression and knowledge transfer;
Ballot Length in Instant Runoff Voting
Kiran Tomlinson (Cornell University), Jon Kleinberg (Cornell University)
Tabular
🎯 What it does: This study investigates how the length of the ballot (i.e., the number of candidates each voter can rank) affects the election results in Instant Runoff Voting (IRV). It proposes that theoretically, up to k-1 different winners can occur and provides a construction for the matching lower bound. It also explores scenarios under preference constraints such as single-peaked and single-crossing. The impact of different ballot lengths on the winners is validated through simulations and actual election data from PrefLib.
Bayesian Cross-Modal Alignment Learning for Few-Shot Out-of-Distribution Generalization
Lin Zhu (Shanghai Jiao Tong University), Nanyang Ye (Shanghai Jiao Tong University)
Domain AdaptationContrastive LearningImageText
🎯 What it does: A Bayes-CAL method is proposed, utilizing Bayesian cross-modal alignment learning to achieve OoD generalization under few samples;
Bayesian Federated Neural Matching That Completes Full Information
Peng Xiao (Tongji University), Samuel Cheng (University of Oklahoma)
ClassificationSegmentationFederated LearningConvolutional Neural NetworkImage
🎯 What it does: A new federated neural matching method (NAFI) is proposed, which compensates for the lack of global information in the matching cost of PFNM by adding a Kullback–Leibler (KL) penalty term to the original PFNM framework.
Bayesian Optimization-Based Combinatorial Assignment
Jakob Weissteiner (University of Zurich), Sven Seuken (University of Zurich)
OptimizationTabularBenchmark
🎯 What it does: A combination allocation mechanism based on Bayesian optimization, BOCA, is proposed to efficiently collect agent preference information in combinatorial auctions.
Beam Search Optimized Batch Bayesian Active Learning
Jingyu Sun (NTT Computer and Data Science Laboratories), Susumu Takeuchi (NTT Computer and Data Science Laboratories)
ClassificationOptimizationComputational EfficiencyConvolutional Neural NetworkImage
🎯 What it does: This paper proposes a diversified beam search method with adaptive constraints applied in batch active learning to select informative and diverse samples in each sampling batch.
Behavior Estimation from Multi-Source Data for Offline Reinforcement Learning
Guoxi Zhang (Kyoto University), Hisashi Kashima (RIKEN)
Reinforcement LearningTabular
🎯 What it does: This study investigates behavior estimation under multi-source data, proposing a latent variable model and the LBRAC-v method to learn and allocate multiple behavior strategies.
Behavioral Learning in Security Games: Threat of Multi-Step Manipulative Attacks
Thanh H. Nguyen (University of Oregon), Arunesh Sinha (Rutgers University)
OptimizationAdversarial AttackReinforcement LearningTabular
🎯 What it does: This study investigates misleading attacks by attackers on learning-based defense strategies in multi-step security games and proposes a corresponding attack plan optimization algorithm.
Bellman Meets Hawkes: Model-Based Reinforcement Learning via Temporal Point Processes
Chao Qu (Ant Group), Hongyuan Mei (Toyota Technological Institute at Chicago)
Reinforcement LearningSequential
🎯 What it does: This paper proposes a modeling reinforcement learning framework based on the Hawkes point process (NHPI) for learning optimal intervention strategies in continuous-time, random event-driven semi-Markov decision problems.
BERT-ERC: Fine-Tuning BERT Is Enough for Emotion Recognition in Conversation
Xiangyu Qin (Peking University), Jinshi Cui (Peking University)
RecognitionKnowledge DistillationTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: This paper proposes a new Emotion Recognition Conversation (ERC) paradigm that fine-tunes fine-grained context and dialogue structure directly on the BERT pre-trained model, constructing the BERT-ERC model.
Bespoke: A Block-Level Neural Network Optimization Framework for Low-Cost Deployment
Jong-Ryul Lee (Electronics and Telecommunications Research Institute), Yong-Hyuk Moon (University of Science and Technology)
ClassificationCompressionOptimizationKnowledge DistillationNeural Architecture SearchConvolutional Neural NetworkImage
🎯 What it does: A low-cost model optimization framework called Bespoke is proposed, which gradually replaces sub-blocks of the original network by randomly sampling sub-network blocks from a teacher model and publicly pre-trained networks and performing knowledge distillation to obtain lightweight models that meet various deployment requirements.
BEST: BERT Pre-training for Sign Language Recognition with Coupling Tokenization
Weichao Zhao (University of Science and Technology of China), Houqiang Li (Huawei Inc.)
RecognitionPose EstimationGraph Neural NetworkTransformerLarge Language ModelVideo
🎯 What it does: This study proposes a BERT pre-trained pose triplet (hand, hand, body) model for sign language recognition.
BETA-CD: A Bayesian Meta-Learned Cognitive Diagnosis Framework for Personalized Learning
Haoyang Bi (University of Science and Technology of China), Jinze Wu (iFLYTEK AI Research)
Meta LearningTabular
🎯 What it does: The BETA-CD framework is proposed, which combines Bayesian hierarchical models and meta-learning for personalized learning cognitive diagnosis, allowing for rapid adaptation to new students and quantifying the uncertainty of diagnostic results.
Better and Faster: Adaptive Event Conversion for Event-Based Object Detection
Yansong Peng (University of Science and Technology of China), Feng Wu (University of Science and Technology of China)
Object DetectionImage
🎯 What it does: This paper proposes the Hyper Histogram representation, Adaptive Event Conversion (AEC) module, and Shadow Mosaic augmentation, integrating them into YOLOv5, Deformable-DETR, and RetinaNet to achieve end-to-end efficient object detection with event cameras.
Better Context Makes Better Code Language Models: A Case Study on Function Call Argument Completion
Hengzhi Pei (University of Illinois Urbana-Champaign), George Karypis (Amazon Web Services)
AI Code AssistantTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: A reproducible collection of Python project environments (PYENVS) was constructed, and based on this, a function call parameter completion dataset (CALLARGS) was created. A program analyzer was used to extract function implementation and usage information across files and projects, studying its impact on the completion performance of pre-trained code language models.
Better Generalized Few-Shot Learning Even without Base Data
Seong-Woong Kim (Inha University), Dong-Wan Choi (Inha University)
ClassificationMeta LearningImage
🎯 What it does: This study proposes a zero-shot generalized few-shot learning method to address the issue of integrating new category knowledge into pre-trained models without base class samples.
Better Peer Grading through Bayesian Inference
Hedayat Zarkoob (University of British Columbia), Kevin Leyton-Brown (University of British Columbia)
Tabular
🎯 What it does: This study improves the Bayesian inference framework of the peer assessment system by incorporating low-effort scoring modeling, upper truncation handling of discrete rating levels, and interpretable weighted average score output.
BEVDepth: Acquisition of Reliable Depth for Multi-View 3D Object Detection
Yinhao Li (Institute of Computing Technology, Chinese Academy of Sciences), Zeming Li (MEGVII Technology)
Object DetectionDepth EstimationAutonomous DrivingTransformerSupervised Fine-TuningPoint Cloud
🎯 What it does: This paper proposes BEVDepth, a multi-view BEV 3D object detection framework achieved through explicit depth supervision and a camera-aware depth prediction module.
BEVStereo: Enhancing Depth Estimation in Multi-View 3D Object Detection with Temporal Stereo
Yinhao Li (University of Chinese Academy of Sciences), Zeming Li (MEGVII Technology)
Object DetectionDepth EstimationAutonomous DrivingPoint Cloud
🎯 What it does: A multi-view 3D object detection framework named BEVStereo is proposed, which enhances depth estimation through dynamic temporal stereo technology and achieves high-precision detection of targets such as vehicles and pedestrians by combining the BEVDepth structure.
Beyond ADMM: A Unified Client-Variance-Reduced Adaptive Federated Learning Framework
Shuai Wang (Information Systems Technology and Design), Defeng Sun (Applied Mathematics)
OptimizationFederated LearningImage
🎯 What it does: The FedVRA framework is proposed, which implements an adaptive federated learning algorithm with reduced client variance by combining the ADMM concept.
Beyond Graph Convolutional Network: An Interpretable Regularizer-Centered Optimization Framework
Shiping Wang (Fuzhou University), Yong Chen (Beijing University of Posts and Telecommunications)
OptimizationExplainability and InterpretabilityGraph Neural NetworkGraph
🎯 What it does: An interpretable optimization framework based on regularization is proposed, and under this framework, a dual-regularization graph convolutional network tsGCN is designed.
Beyond Smoothing: Unsupervised Graph Representation Learning with Edge Heterophily Discriminating
Yixin Liu (Monash University), Shirui Pan (Griffith University)
Representation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: This paper proposes an unsupervised graph representation learning framework called GREET, which can distinguish between similar edges and dissimilar edges, and generates node representations using both types of edges.
Bi-directional Feature Reconstruction Network for Fine-Grained Few-Shot Image Classification
Jijie Wu (Lanzhou University of Technology), Yi-Zhe Song (University of Surrey)
ClassificationMeta LearningTransformerImage
🎯 What it does: A bidirectional feature reconstruction network (Bi-FRN) is proposed, achieving bidirectional reconstruction from support to query and from query to support, enhancing the discriminative ability of fine-grained few-shot classification.
Bidding Graph Games with Partially-Observable Budgets
Guy Avni (University of Haifa), Đorđe Žikelić (Institute of Science and Technology Austria)
🎯 What it does: This paper proposes and analyzes two-player zero-sum games with partially observable budgets (partial-information bidding games), focusing particularly on the Poorman bidding mechanism under the mean payoff objective. It provides the optimal pure strategy and value expressions for partially informed players, revealing the phenomenon that value may not exist.
Bidirectional Domain Mixup for Domain Adaptive Semantic Segmentation
Daehan Kim (Hanbat National University), Dong-Geol Choi (Hanbat National University)
SegmentationDomain AdaptationImage
🎯 What it does: A Bidirectional Domain Mixup method is proposed for unsupervised semantic segmentation domain adaptation.
Bidirectional Optical Flow NeRF: High Accuracy and High Quality under Fewer Views
Shuo Chen (Beijing University of Posts and Telecommunications), Huaming Wan (Beijing University of Posts and Telecommunications)
GenerationData SynthesisPose EstimationNeural Radiance FieldOptical FlowImage
🎯 What it does: This paper studies the improvement of geometric estimation and view synthesis of NeRF using bidirectional optical flow supervision from a sparse perspective, proposing a bidirectional optical flow NeRF and view enhancement fusion method.
Bilinear Exponential Family of MDPs: Frequentist Regret Bound with Tractable Exploration & Planning
Reda Ouhamma (University of Lille), Odalric Maillard (University of Lille)
Reinforcement Learning
🎯 What it does: This paper proposes a BEF-RLSVI algorithm for episodic reinforcement learning problems with continuous state-action spaces, where rewards and transitions are unknown.
Black-Box Adversarial Attack on Time Series Classification
Daizong Ding (Fudan University), Min Yang (Fudan University)
ClassificationAdversarial AttackTime Series
🎯 What it does: This paper proposes a black-box adversarial attack method for time series classification.
Boosted Dynamic Neural Networks
Haichao Yu (University of Illinois Urbana-Champaign), Humphrey Shi (Wormpex AI Research)
ClassificationImage
🎯 What it does: Designed and trained BoostNet, a framework that treats early exit dynamic neural networks as a gradient boosting additive model to address the issue of data distribution mismatch between the training and inference phases.
Boosting Few-Shot Text Classification via Distribution Estimation
Han Liu (Dalian University of Technology), Xianchao Zhang (Dalian University of Technology)
ClassificationMeta LearningTransformerSupervised Fine-TuningText
🎯 What it does: Two query sample-based distribution estimation strategies (Way-DE and Shot-DE) are proposed, which use these estimated Gaussian distributions to generate additional samples to enhance few-shot text classification models.
Boosting Graph Neural Networks via Adaptive Knowledge Distillation
Zhichun Guo (University of Notre Dame), Nitesh V. Chawla (University of Notre Dame)
Knowledge DistillationGraph Neural NetworkGraph
🎯 What it does: This paper proposes the BGNN framework, which utilizes sequential knowledge distillation combined with adaptive temperature and misclassified sample weighting methods to enhance the performance of a single GNN by leveraging complementary knowledge from different GNN models.
Boosting Point Clouds Rendering via Radiance Mapping
Xiaoyang Huang (Shanghai Jiao Tong University), Wenjun Zhang (Anhui University)
GenerationComputational EfficiencyNeural Radiance FieldPoint Cloud
🎯 What it does: An efficient rendering method based on sparse point clouds is proposed, utilizing radiance mapping to achieve single-pixel inference.
Boosting Semi-Supervised Semantic Segmentation with Probabilistic Representations
Haoyu Xie (Northeastern University), Chang Xu (University of Sydney)
SegmentationContrastive LearningImage
🎯 What it does: This paper proposes a probabilistic pixel representation contrastive learning framework (PRCL), which maps pixels to multivariate Gaussian distributions. It utilizes probabilistic information to suppress the negative impact of inaccurate pseudo-labels on contrastive learning, thereby enhancing the representation quality of semi-supervised semantic segmentation.
Bootstrapping Multi-View Representations for Fake News Detection
Qichao Ying (Fudan University), Shiming Ge (Chinese Academy of Sciences)
ClassificationAnomaly DetectionRepresentation LearningConvolutional Neural NetworkTransformerMixture of ExpertsImageTextMultimodality
🎯 What it does: The Bootstrapping Multi-View Representations (BMR) method is proposed, which extracts multi-modal representations using three views (text, image modality, image semantics) and achieves feature refinement and fusion through an improved MMoE. Subsequently, adaptive re-weighting and bootstrapping of representations are performed using single-view prediction and cross-modal consistency learning to enhance fake news detection performance.
Boundary Graph Neural Networks for 3D Simulations
Andreas Mayr (Johannes Kepler University Linz), Johannes Brandstetter (Johannes Kepler University Linz)
Graph Neural NetworkGraphPhysics Related
🎯 What it does: An effective theory is proposed, and based on this, Boundary Graph Neural Networks (BGNNs) are constructed to dynamically represent complex geometric boundaries (such as hoppers, rotating drums, and mixers) in three-dimensional granular flow simulations, achieving efficient modeling of particle-boundary interactions.
Breaking Immutable: Information-Coupled Prototype Elaboration for Few-Shot Object Detection
Xiaonan Lu (Aerospace Information Research Institute, Chinese Academy of Sciences), Kun Fu (Aerospace Information Research Institute, Chinese Academy of Sciences)
Object DetectionConvolutional Neural NetworkImage
🎯 What it does: A Few-Shot object detection framework ICPE based on conditional information coupling and dynamic aggregation is proposed, which can generate high-quality prototypes with query awareness for each query image.
BridgeTower: Building Bridges between Encoders in Vision-Language Representation Learning
Xiao Xu (Harbin Institute of Technology), Nan Duan (Microsoft Research Asia)
Representation LearningTransformerVision Language ModelImageTextMultimodality
🎯 What it does: The BridgeTower model is proposed, which connects the top layers of visual and text encoders to each layer of the cross-modal encoder through multiple bridge layers, achieving alignment and fusion at multiple semantic levels.
C-NTPP: Learning Cluster-Aware Neural Temporal Point Process
Fangyu Ding (Shanghai Jiao Tong University), Haiyang Wang (Ant Group)
TransformerAuto EncoderTime SeriesSequentialBiomedical DataFinance Related
🎯 What it does: This paper proposes a clustering-aware neural temporal point process (c-NTPP) based on variational autoencoders, which can automatically infer hidden clusters in event sequences and improve event representation using a clustering attention mechanism.
Calibrated Teacher for Sparsely Annotated Object Detection
Haohan Wang (Shenzhen International Graduate School Tsinghua University), Haoqian Wang (Shenzhen International Graduate School Tsinghua University)
Object DetectionConvolutional Neural NetworkImage
🎯 What it does: The Calibrated Teacher framework is proposed, which generates more reliable pseudo-labels by online calibrating the confidence of the teacher network in sparse annotated object detection, and combines Focal IoU Weight to reduce the misguidance of negative samples caused by missing annotations.
CALIP: Zero-Shot Enhancement of CLIP with Parameter-Free Attention
Ziyu Guo (Peking University), Bin Cui (Carnegie Mellon University)
ClassificationRecognitionTransformerContrastive LearningImagePoint Cloud
🎯 What it does: This paper proposes the CALIP method, which achieves cross-modal interaction between visual and textual features in CLIP through a non-parametric attention mechanism to enhance zero-shot classification performance.
Can Bad Teaching Induce Forgetting? Unlearning in Deep Networks Using an Incompetent Teacher
Vikram S Chundawat (Mavvex Labs), Mohan Kankanhalli (National University of Singapore)
Computational EfficiencyKnowledge DistillationAdversarial AttackConvolutional Neural NetworkRecurrent Neural NetworkTransformerImage
🎯 What it does: This paper proposes a deep network machine forgetting method based on a teacher-student framework, utilizing skilled teachers (fully trained models) and unskilled teachers (randomly initialized or partially trained models) for selective knowledge transfer to students, thereby achieving effective 'forgetting' of specified samples or categories.
Can Label-Specific Features Help Partial-Label Learning?
Ruo-Jing Dong (Southeast University), Min-Ling Zhang (Southeast University)
ClassificationGraph Neural NetworkSupervised Fine-TuningTabular
🎯 What it does: This paper proposes a label-specific feature-based unsupervised learning method called UCL, which first generates soft pseudo-labels through graph label enhancement, then classifies samples into positive, negative, and uncertain based on confidence levels, utilizes spectral clustering to obtain cluster centers to construct label-specific features, and finally combines these with existing classifiers (such as SURE) to improve performance.
Can We Find Strong Lottery Tickets in Generative Models?
Sangyeop Yeo (Ulsan National Institute of Science and Technology), Jaejun Yoo (Ulsan National Institute of Science and Technology)
GenerationData SynthesisGenerative Adversarial NetworkImage
🎯 What it does: This paper studies the search for Strong Lottery Tickets (SLT) in generative models, which can achieve or exceed the generative performance of the full model in a sparse subnetwork without updating weights at random initialization.
CasFusionNet: A Cascaded Network for Point Cloud Semantic Scene Completion by Dense Feature Fusion
Jinfeng Xu (Huazhong University of Science and Technology), Min Chen (Huazhong University of Science and Technology)
RestorationSegmentationGraph Neural NetworkTransformerPoint Cloud
🎯 What it does: A cascade network named CasFusionNet is proposed, which utilizes dense feature fusion to achieve joint inference of point cloud semantic scene completion and semantic segmentation.
Causal Conditional Hidden Markov Model for Multimodal Traffic Prediction
Yu Zhao (Beihang University), Mulan Wang (Beihang University)
Recurrent Neural NetworkGraph Neural NetworkAuto EncoderMultimodalityTime Series
🎯 What it does: This paper proposes the Causal Conditional Hidden Markov Model (CCHMM), which views multimodal traffic flow prediction as a conditional Markov process and decouples the core physical concepts of the traffic system (regional attraction factors, traffic demand factors, and speed factors) through causal structures.
Causal Effect Identification in Cluster DAGs
Tara V. Anand (Columbia University), Elias Bareinboim (Columbia University)
🎯 What it does: Proposes 'Cluster DAG (C-DAG)' as a new graphical model that allows researchers to provide only coarse-grained causal relationships between sets of variables without needing to fully specify all relationships among the variables; uses C-DAG for probabilistic, intervention, and counterfactual inference on macro variables;
Causal Inference with Conditional Instruments Using Deep Generative Models
Debo Cheng (Guangxi Normal University), Thuc Duy Le (University of South Australia)
Representation LearningAuto EncoderTabularFinance Related
🎯 What it does: A conditional instrumental variable (CIV) method based on deep generative models is proposed to discover CIVs and their conditional sets from data with potential confounding factors, in order to estimate average causal effects.
Causal Intervention for Human Trajectory Prediction with Cross Attention Mechanism
Chunjiang Ge (Tsinghua University), Gao Huang (Tsinghua University)
Recurrent Neural NetworkGraph Neural NetworkTransformerTime SeriesSequential
🎯 What it does: By constructing a structural causal model and treating the social environment as a confounding factor, a Social Environment Adjustment (SEAD) method is proposed to intervene in the causal relationships in human trajectory prediction, eliminating the spurious correlations caused by the social environment.
Causal Recurrent Variational Autoencoder for Medical Time Series Generation
Hongming Li (University of Florida), Jose Principe (University of Florida)
GenerationData SynthesisRecurrent Neural NetworkAuto EncoderTime SeriesBiomedical Data
🎯 What it does: Developed a cyclic variational autoencoder (CR-VAE) integrated with Granger causality graphs for the generation and causal discovery of medical time series.
Causes of Stability in Dynamic Coalition Formation
Niclas Boehmer (Technische Universität Berlin), Anna Maria Kerkmann (Heinrich-Heine-Universität Düsseldorf)
🎯 What it does: This paper studies the collective action process in variable utility Cardian Goodell games, based on two dynamic utility change mechanisms: 'resentment' and 'appreciation'. It provides a theoretical analysis of convergence under different stability concepts and experimental validation.
CCQ: Cross-Class Query Network for Partially Labeled Organ Segmentation
Xuyang Liu (ShanghaiTech University), Sibei Yang (ShanghaiTech University)
SegmentationConvolutional Neural NetworkTransformerImageComputed Tomography
🎯 What it does: A cross-category query network (CCQ) is proposed, which learns cross-category semantic concepts of different organs and combines attention mechanisms to train a single multi-organ/tumor segmentation model from multiple partially labeled datasets.
CDMA: A Practical Cross-Device Federated Learning Algorithm for General Minimax Problems
Jiahao Xie (Zhejiang University), Hui Qian (Zhejiang University)
OptimizationFederated LearningGenerative Adversarial NetworkImage
🎯 What it does: A general minimax problem algorithm for cross-device federated learning, CDMA, is proposed;
CDTA: A Cross-Domain Transfer-Based Attack with Contrastive Learning
Zihan Li (Sun Yat-sen University), Michael R. Lyu (Chinese University of Hong Kong)
Domain AdaptationAdversarial AttackContrastive LearningImage
🎯 What it does: A cross-domain transferable attack method CDTA is designed, utilizing unlabeled contrastive learning to train a feature extractor and generate attack samples.
CEE-Net: Complementary End-to-End Network for 3D Human Pose Generation and Estimation
Haolun Li (University of Macau), Chi-Man Pun (University of Macau)
GenerationPose EstimationGenerative Adversarial NetworkImage
🎯 What it does: A complementary end-to-end network CEE-Net is proposed, which jointly trains a 3D pose generator and estimator to improve the accuracy of 3D pose estimation across datasets and for sparse complex poses.
CEM: Constrained Entropy Maximization for Task-Agnostic Safe Exploration
Qisong Yang (Delft University of Technology), Matthijs T.J. Spaan
Safty and PrivacyRobotic IntelligenceReinforcement Learning
🎯 What it does: The research focuses on safe and efficient exploration in the context of unknown tasks, proposing the CEM algorithm to achieve state entropy maximization under safety constraints.
CEMA – Cost-Efficient Machine-Assisted Document Annotations
Guowen Yuan (University of Hong Kong), Tien-Hsuan Wu (University of Hong Kong)
OptimizationData-Centric LearningTransformerSupervised Fine-TuningText
🎯 What it does: With a limited budget, a machine learning-assisted approach is designed to help humans complete complex document annotations, along with a cost assessment and knowledge estimation document selection strategy.
Centerless Multi-View K-means Based on the Adjacency Matrix
Han Lu (Xidian University), Wei Xia (University of Evansville)
OptimizationGraph Neural NetworkMultimodality
🎯 What it does: This paper proposes a centralized multi-view K-Means clustering method that constructs a distance matrix using an adjacency matrix and fully exploits multi-view information through tensor Schatten p-norm regularization, completing clustering without the need for initializing centroids.
CertiFair: A Framework for Certified Global Fairness of Neural Networks
Haitham Khedr (University of California), Yasser Shoukry (University of California)
ClassificationOptimizationTabular
🎯 What it does: A verifier has been constructed to prove global individual fairness for neural networks (NN), and a fairness regularization training method based on product networks has been proposed, enabling the NN to provide the same classification results when faced with similar individuals.
Certifiable Out-of-Distribution Generalization
Nanyang Ye (Shanghai Jiao Tong University), Jun Zhu (Tsinghua University)
ClassificationDomain AdaptationAnomaly DetectionImageBenchmark
🎯 What it does: A certified outlier distribution generalization method based on stochastic perturbation and maximum margin training (SDL) is proposed, which provides theoretical accuracy guarantees under different types of distribution shifts.
Certifying Fairness of Probabilistic Circuits
Nikil Roashan Selvam (University of California), YooJung Choi (Arizona State University)
OptimizationTabular
🎯 What it does: This study investigates the fairness issues under partial observation conditions in probabilistic circuits (PC), proposes and tests the definition of discrimination patterns, and provides a framework for summarizing these patterns.
CF-ViT: A General Coarse-to-Fine Method for Vision Transformer
Mengzhao Chen (Xiamen University), Rongrong Ji (Xiamen University)
ClassificationTransformerImage
🎯 What it does: CF-ViT is proposed, a two-stage coarse-fine granularity visual Transformer that first uses coarse granularity partitioning for rapid classification, and if confidence is insufficient, it performs fine-grained re-partitioning on important areas to further improve accuracy.
CFFT-GAN: Cross-Domain Feature Fusion Transformer for Exemplar-Based Image Translation
Tianxiang Ma (University of Chinese Academy of Sciences), Tieniu Tan (Chinese Academy of Sciences)
Image TranslationGenerationConvolutional Neural NetworkTransformerGenerative Adversarial NetworkImage
🎯 What it does: The CFFT-GAN framework is proposed, utilizing Cross-Domain Feature Fusion Transformer (CFFT) to achieve high-quality style transfer in exemplar-based image translation tasks.
Channel Regeneration: Improving Channel Utilization for Compact DNNs
Ankit Sharma (University of Central Florida), Hassan Foroosh (University of Central Florida)
ClassificationOptimizationConvolutional Neural NetworkImage
🎯 What it does: A new channel regeneration technology is proposed, which activates redundant channels in deep neural networks by reinitializing the scale factor γ of batch normalization, thereby improving channel utilization.
Characterizing Structural Hardness of Logic Programs: What Makes Cycles and Reachability Hard for Treewidth?
Markus Hecher (Massachusetts Institute of Technology)
🎯 What it does: A tree-width-aware reduction from SAT to normal ASP using reachability and cycles is proposed, reducing tree-width from k to O(k log k), and it is proven that this reduction is optimal under the ETH hypothesis;
Circuit Minimization with QBF-Based Exact Synthesis
Franz-Xaver Reichl (TU Wien), Stefan Szeider (TU Wien)
OptimizationTabularBenchmark
🎯 What it does: A precise synthesis and local rewriting method based on QBF is designed to minimize subcircuits in large circuits.
CL3D: Unsupervised Domain Adaptation for Cross-LiDAR 3D Detection
Xidong Peng (ShanghaiTech University), Yuexin Ma (ShanghaiTech University)
Object DetectionDomain AdaptationAutonomous DrivingPoint Cloud
🎯 What it does: Proposes CL3D, an unsupervised domain adaptation method for cross-LiDAR 3D detection;
Class Fairness in Online Matching
Hadi Hosseini (Pennsylvania State University), Nisarg Shah (University of Toronto)
Optimization
🎯 What it does: In online bipartite matching, fairness among different types of agents is considered, and two deterministic algorithms—MATCH-AND-SHIFT and EQUAL-FILLING—are proposed for divisible and indivisible items to achieve approximate guarantees for class-level EF1, MMS, CPROP, and efficiency USW.
Class Overwhelms: Mutual Conditional Blended-Target Domain Adaptation
Pengcheng Xu (Western University), Charles Ling (Western University)
ClassificationDomain AdaptationGenerative Adversarial NetworkImage
🎯 What it does: To address the problem of single-source multi-target hybrid domain adaptation (BTDA), a mutual conditional alignment framework (MCDA) is proposed, which jointly aligns the class distribution P(Z|Y) and the classifier P(Y|Z) without requiring domain labels, and enhances style using low-level features to correct biased classifiers.
Class-Independent Regularization for Learning with Noisy Labels
Rumeng Yi (Beijing Jiaotong University), Shijian Lu (Nanyang Technological University)
ClassificationRecognitionConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: This paper proposes a Class-Independent Regularization (CIR) for training deep networks under noisy labels, which regularizes the traditional softmax classifier by decomposing K-class multiclass classification into K independent binary classifiers, and implements a heterogeneous adaptive co-teaching mechanism to achieve complementary training of sample selection and classification.
ClassFormer: Exploring Class-Aware Dependency with Transformer for Medical Image Segmentation
Huimin Huang (Zhejiang University), Yefeng Zheng (Tencent Jarvis Lab)
SegmentationTransformerImageBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: This paper proposes ClassFormer, a Transformer module for medical image segmentation, specifically addressing two major issues: intra-class consistency at the pixel level and inter-class differentiation.
CLIP-ReID: Exploiting Vision-Language Model for Image Re-identification without Concrete Text Labels
Siyuan Li (East China Normal University), Qingli Li (East China Normal University)
RecognitionRetrievalTransformerVision Language ModelContrastive LearningImage
🎯 What it does: Fine-tune the pre-trained CLIP vision-language model for person/vehicle ReID tasks and propose a two-stage training strategy to introduce learnable text tokens.
CLIPVG: Text-Guided Image Manipulation Using Differentiable Vector Graphics
Yiren Song (Shanghai Jiao Tong University), Minzhe Li (Shanghai Jiao Tong University)
Image TranslationOptimizationTransformerContrastive LearningImage
🎯 What it does: Introducing CLIPVG: a text-guided image editing framework based on differentiable vector graphics, which optimizes color and geometric parameters directly in the vector domain using CLIP loss, enabling semantic image editing without the need for additional generative models.
Cluster-Guided Contrastive Graph Clustering Network
Xihong Yang (National University of Defense Technology), En Zhu (Nanjing University of Aeronautics and Astronautics)
ClassificationRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: A Cluster-guided Contrastive Graph Clustering network (CCGC) is proposed, which enhances the performance of unsupervised graph node clustering by guiding the construction of positive and negative samples through high-confidence clustering results.
Clustering What Matters: Optimal Approximation for Clustering with Outliers
Akanksha Agrawal (Indian Institute of Technology Madras), Jie Xue (New York University Shanghai)
Optimization
🎯 What it does: A general reduction framework is proposed, transforming clustering problems with outliers (such as k-MEDIANOUT, k-MEANSOUT, etc.) into standard clustering problems without outliers while preserving the approximation ratio, thus obtaining a parameterized FPT approximation algorithm.