Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks142 from public data
- 1,836
- A Comprehensive Survey of Continual Learning: Theory, Method and Application1,672
This work presents a comprehensive survey of continual learning, seeking to bridge the basic settings, theoretical foundations, representative methods, and practical applications, and summarizes the general objectives of continual learning as ensuring a proper stability-plasticity trade-off and an adequate intra/inter-task generalizability in the context of resource efficiency.
- 1,075
- Cross-domain collaboration recommendation351
The Cross-domain Topic Learning (CTL) model is proposed, which consolidates the existing cross-domain collaborations through topic layers instead of at author layers, which alleviates the sparseness issue and outperforms baselines significantly on multiple recommendation metrics.
- 289
- Benchmarking Robustness of 3D Object Detection to Common Corruptions in Autonomous Driving228
Several important findings are drawn, including: 1) motion-level corruptions are the most threatening ones that lead to significant performance drop of all models; 2) LiDAR-camerafusion models demonstrate better robustness; 3) camera-only models are extremely vulnerable to image corruptions, showing the indispensability of LiDar point clouds.
- Pruning from Scratch225
This work finds that pre-training an over-parameterized model is not necessary for obtaining the target pruned structure, and empirically shows that more diverse pruned structures can be directly pruned from randomly initialized weights, including potential models with better performance.
- 215
- Toward the third generation artificial intelligence209
This paper looks toward developing a third generation artificial intelligence by combining the current paradigms, and considers symbolism as the first generation of AI and connectionism as the second generation.
- A Comprehensive Study on Robustness of Image Classification Models: Benchmarking and Rethinking187
A comprehensive robustness benchmark called ARES-Bench is established on the image classification task and several important findings are drawn, including that there exists an intrinsic trade-off between the adversarial and natural robustness of specific noise types for the same model architecture.
- Text-to-Image Diffusion Models can be Easily Backdoored through Multimodal Data Poisoning149
This work performs a systematic investigation of backdoor attack on text-to-image diffusion models and proposes BadT2I, a general multimodal backdoor attack framework that tampers with image synthesis in diverse semantic levels.
- Improving Interpretability of Deep Neural Networks with Semantic Information134
This work proposes a novel technique to improve the interpretability of DNNs by leveraging the rich semantic information embedded in human descriptions by extracting a set of semantically meaningful topics from the human descriptions that cover a wide range of visual concepts.
- Learning to Write Stylized Chinese Characters by Reading a Handful of Examples114
A novel framework of Style-Aware Variational Auto-Encoder (SA-VAE), which disentangles the content-relevant and style-relevant components of a Chinese character feature with a novel intercross pair-wise optimization method and has a powerful one-shot/few-shot generalization ability.
- 111
- Towards Effective Adversarial Textured 3D Meshes on Physical Face Recognition75
The goal of this work is to develop a more reliable technique that can carry out an end-to-end evaluation of adversarial robustness for commercial systems, and designs adversarial textured 3D meshes with an elaborate topology on a human face, which can be 3D-printed and pasted on the attacker's face to evade the defenses.
- Detection Transformer with Stable Matching73
This paper proposes two simple yet effective modifications by integrating positional metrics to DETR’s classification loss and matching cost, named position-supervised loss and position-modulated cost, and shows consistent improvements over baselines.
- Multi-task learning for segmentation and classification of breast tumors from ultrasound images70
A novel multi-task learning network designed to optimize tumor segmentation and classification in BUS images, which not only outperforms mainstream multi-task learning methods for both breast tumor segmentation and classification tasks, but also achieves state-of-the-art results for BUS tumor segmentation.
- 69
- Ensemble of kernel extreme learning machine based random forest classifiers for automatic heartbeat classification67
An automatic ECG heartbeat classification method based on ensemble learning and multi-kernel learning is proposed, which achieves a superior classification performance compared to the state-of-the-art methods.
- Triple-Memory Networks: A Brain-Inspired Method for Continual Learning66
A novel approach named triple memory networks (TMNs) for continual learning that model the interplay of hippocampus, prefrontal cortex and sensory cortex as a triple-network architecture of generative adversarial networks (GAN) and achieves new state-of-the-art performance on a variety of class-incremental learning benchmarks.
- Exploring the Robustness of Decision-Level Through Adversarial Attacks on LLM-Based Embodied Models63
The Embodied Intelligent Robot Attack Dataset (EIRAD) is constructed, tailored specifically for robustness evaluation, and experimental results demonstrate that the method exhibits a superior attack success rate when targeting LLM-based embodied models, indicating a lower level of decision-level robustness in these models.
- Towards Transferable Targeted 3D Adversarial Attack in the Physical World48
A novel framework named TT3D is designed that could rapidly reconstruct from few multi-view images into Transferable Targeted 3D textured meshes and exhibits superior cross-model transferability but also maintains considerable adaptability across different renders and vision tasks.
- Unsupervised Part Segmentation through Disentangling Appearance and Shape42
A bottleneck block is designed to squeeze and expand the appearance representation, leading to a more effective disentanglement between geometry and appearance, and combined with a self-supervised part classification loss and an improved geometry concentration constraint can segment more consistent parts with semantic meanings.
- 38
- 37
- 37
- Playing FPS Games With Environment-Aware Hierarchical Reinforcement Learning36
This work presents a hierarchical model that works in a manager-worker fashion over two levels of hierarchy that significantly outperforms the alternative RL-based models on FPS games requiring maze solving and combat skills, etc.
- HiDe-PET: Continual Learning via Hierarchical Decomposition of Parameter-Efficient Tuning33
Hierarchical Decomposition PET (HiDe-PET), an innovative approach that explicitly optimizes the decomposed objective through incorporating task-specific and task-shared knowledge via mainstream PET techniques along with efficient recovery of pre-trained representations, demonstrates remarkably superior performance over a broad spectrum of recent strong baselines.
- 32
- 31
- 30
- Exploring Transferability of Multimodal Adversarial Samples for Vision-Language Pre-Training Models With Contrastive Learning28
A novel gradient-based multimodal adversarial attack method, underpinned by contrastive learning, to improve the transferability of multimodal adversarial samples in VLP models, concurrently generates adversarial texts and images within imperceptive perturbation.
- AnyPos: Automated Task-Agnostic Actions for Bimanual Manipulation28
AnyPos, a unified pipeline that integrates large-scale automated task-agnostic exploration with robust embodiment modeling through inverse dynamics learning, is introduced, a unified pipeline that integrates large-scale automated task-agnostic exploration with robust embodiment modeling through inverse dynamics learning.
- Controllable Navigation Instruction Generation with Chain of Thought Prompting27
C-Instructor is proposed, which utilizes the chain-of-thought-style prompt for style-controllable and content-controllable instruction generation and introduces a Style-Mixed Training policy, harnessing the prior knowledge of LLMs to enable style control for instruction generation based on different prompts within a single model instance.
- Exploring the Transferability of Visual Prompting for Multimodal Large Language Models26
This paper proposes Transferable Visual Prompting (TVP), a simple and effective approach to generate visual prompts that can transfer to different models and improve their performance on downstream tasks after trained on only one model.
- Robust and Efficient Graph Correspondence Transfer for Person Re-Identification26
A novel pose context descriptor to accurately model human body configurations, and an approach to measure the similarity between a pair of pose context descriptors are presented, which can effectively and efficiently handle the spatial misalignment problem in Re-ID.
- 26
- Joint dictionary and metric learning for person re-identification26
This paper proposes to formulate robust feature representation learning and discriminative metric learning into a unified framework for person re-identification, and to alleviate the issue of amount bias towards hard negative pairs in metric learning.
- 26
- Graph Correspondence Transfer for Person Re-Identification25
The GCT model formulates person re-identification as an off-line graph matching and on-line correspondence transferring problem and can handle spatial misalignment caused by large variations in view angles and human poses owing to the benefits of patch-wise graph matching.
- HybridDepth: Robust Metric Depth Fusion by Leveraging Depth from Focus and Single-Image Priors23
Comprehensive quantitative and qualitative analyses demonstrate that Hybriddepth outperforms state-of-the-art (SOTA) models on common datasets such as DDFF12 and NYU Depth V2, and shows strong zero-shot generalization.
- 23
- Stochastic Quantization for Learning Accurate Low-Bit Deep Neural Networks23
Experiments show that SQ can consistently and significantly improve the accuracy for different low-bit DNNs on various datasets and various network structures, no matter whether activation values are quantized or not.
- Learning Task-Distribution Reward Shaping with Meta-Learning21
This paper considers reward shaping on a distribution of tasks that share state spaces but not necessarily action spaces, and proposes a novel meta-learning framework to automatically learn such reward shaping to apply on newly sampled tasks.
- Weighted bilinear coding over salient body parts for person re-identification20
This work proposes a novel weighted bilinear coding (WBC) framework for local feature aggregation in CNN networks to pursue more representative and discriminative feature representations, which can adapt to other state-of-the-art methods and improve their performance.
- Combo-Action: Training Agent For FPS Game with Auxiliary Tasks20
A novel method which can plan on temporally-extended action sequences, which is referred as Combo-Action to compress the action space and outperforms previous state-of-the-art approaches by a large margin.
- Video-to-Video Translation with Global Temporal Consistency20
This paper proposes an unified video-to-video translation framework based on the GAN framework, where a residual error based mechanism is presented to ensure the video-level consistency of the same location in different frames (lobal temporal consistency).
- WFSS: weighted fusion of spectral transformer and spatial self-attention for robust hyperspectral image classification against adversarial attacks19
A weighted fusion of spectral transformer and spatial self-attention (WFSS) is designed to achieve the multi-scale fusion of spectral and spatial connections, which further improves the model’s robustness.
- Training Robust Deep Collaborative Filtering Models via Adversarial Noise Propagation19
A novel adversarial training framework, Random Layer-wise Adversarial Training (RAT), which trains a robust deep collaborative filtering model via adversarial noise propagation in a random layer-wise manner and demonstrates the effectiveness of RAT on multilayer perceptron (MLP) and convolutional neural networks-basedDeep collaborative filtering models.
- SVQN: Sequential Variational Soft Q-Learning Networks19
A novel algorithm is proposed, named sequential variational soft Q-learning networks (SVQNs), which formalizes the inference of hidden states and maximum entropy reinforcement learning (MERL) under a unified graphical model and optimizes the two modules jointly.
- Detecting obstructive sleep apnea by craniofacial image–based deep learning18
The results suggest that combining craniofacial photographs and deep learning techniques can help detect OSA automatically, and the model may have potential utility as a tool to assess OSA probability in clinics or screen for OSA in the community.
- Cell-sensitive phase contrast microscopy imaging by multiple exposures18
A cell-sensitive camera response function which responds to cells' irradiance signals but generates a constant on non-cell background signal is estimated, which reveals the appearance details of cells and suppresses background noise near zero.
- Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment16
Experimental results show that CCA can significantly enhance the guidance-free performance of all tested models with just one epoch of fine-tuning on the pretraining dataset, on par with guided sampling methods, and largely removes the need for guided sampling in AR visual generation and cuts the sampling cost by half.
- Delving into the Adversarial Robustness on Face Recognition.16
A landmark-guided cutout (LGC) attack method is proposed to improve the transferability of adversarial examples for black-box attacks, by considering the special characteristics of face recognition.
- AdvFAS: A robust face anti-spoofing framework against adversarial examples15
A robust face anti-spoofing framework is proposed, namely AdvFAS, that leverages two coupled scores to accurately distinguish between correctly detected and wrongly detected face images.
- Convolutional neural network with adaptive inferential framework for skeleton-based action recognition15
A novel convolutional neural network with adaptive inferential framework (AIF-CNN) to exploit the dependencies among the skeleton joints and demonstrate that the performance is improved significantly by integrating the different prior information.
- Gradient-based conditional generative adversarial network for non-uniform blind deblurring via DenseResNet14
The authors solve blind image deblurring based on a conditional generative adversarial network (GAN), where the sharp image is restored by an end-to-end trainable neural network based on dense blocks and residual network (DenseResNet).
- Natural Image Deblurring Based on Ringing Artifacts Removal via Knowledge-Driven Gradient Distribution Priors12
This work finds that the blur process changes the distribution of the image gradient, and thus attempts to use the priori knowledge for guiding blind deblurring, and proposes a non-blind deconvolution method based on fitted multi-order gradient priors as regularization to restore the latent image.
- 10
- Detection of pulmonary embolism severity using clinical characteristics, hematological indices, and machine learning techniques10
The experiment findings show that the indicators chosen, such as syncope, systolic blood pressure (SBP), oxygen saturation, white blood cell (WBC), neutrophil percentage (NEUT%), and others, are crucial for the feature selection approach presented in this study to assess the severity of PE.
- Combining Tree Search and Action Prediction for State-of-the-Art Performance in DouDiZhu10
This paper extends AlphaZero to multiplayer IIGs by developing a new MCTS method, Action-Prediction MCTS (AP-MCTS), which first builds the search tree based on public information, adopts the policy-value network to generalize between hidden states, and finally predicts other players' actions directly.
- Motion sketch based crowd video retrieval10
This paper proposes to leverage the motion structure coding algorithm for motion-level video indexing and hand-drawn sketch representation and exploit distance metric fusion strategy incorporated with Ranking SVM for measuring the relevant degree between a sketch query and crowd videos.
- 9
- 8
- 8
- 8
- Semi-supervised Max-margin Topic Model with Manifold Posterior Regularization8
This paper presents an effective semi-supervised max-margin topic model by naturally introducing manifold posterior regularization to a regularized Bayesian topic model, named LapMedLDA, which jointly learns latent topics and a related classifier with only a small fraction of labeled documents.
- 7
- 7
- 6
- 6
- Efficient and Robust Semi-supervised Learning Over a Sparse-Regularized Graph6
Experimental results on real datasets validate that the proposed novel GSSL based on a batch of informative beacons with sparsity appropriately harnessed is effective and efficient to implement scalable inference, robust to sample corruptions, and capable to boost the performance incrementally in an open set by updating the beacon-related parameters.
- 5
- 5
- 5
- 5
- 5
- 5
- 4
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- Towards controllable image descriptions with semi-supervised VAE3
This work proposes a semi-supervised deep generative model, Semi- supervised Conditional Variational Auto-Encoder (SCVAE), capable of leveraging more labelled and unlabelled data in theGenerative model schema.
- 3
- 3
- 3
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- Toward Accurate Visual Reasoning With Dual-Path Neural Module Networks2
A novel method of Dual-Path Neural Module Network which can implement complex visual reasoning by forming a more flexible layout regularized by the pairwise loss and significantly improves the performance of neural module networks especially on the real-world datasets.
- 2
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Learn a Robust Policy in Adversarial Games via Playing with an Expert Opponent–
This work technically proposes a novel method called Ladder Opponent Modeling (LOM), which builds a ladder opponent to facilitate the learning process and alleviates the sparse reward issue when facing the expert opponent at the beginning.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-16. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.