Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks59 from public data
- Advancing Image Understanding in Poor Visibility Environments: A Collective Benchmark Study351
The UG2+ challenge Track 2 competition in IEEE CVPR 2019 is launched, aiming to evoke a comprehensive discussion and exploration about whether and how low-level vision techniques can benefit the high-level automatic visual recognition in various scenarios.
- MemOS: A Memory OS for AI System138
MemOS is proposed, a memory operating system that treats memory as a manageable system resource that brings controllability, plasticity, and evolvability to LLMs, laying the foundation for continual learning and personalized modeling.
- On penalty-based bilevel gradient descent method137
This work shows that under certain conditions, the penalty reformulation recovers the (local) solutions of the original bilevel problem and proposes the penalty-based bilevel gradient descent (PBGD) algorithm and establishes its finite-time convergence for the constrained bilevel problem with lower-level constraints yet without lower-level strong convexity.
- 99
- Recurrent Exposure Generation for Low-Light Face Detection95
The proposed method, named REGDet, is the first ‘detection-with-enhancement’ framework for low-light face detection and not only encourages rich interaction and feature fusion across different illumination levels, but also enables effective end-to-end learning of the REG component to be better tailored for face detection.
- LOGER: A Learned Optimizer Towards Generating Efficient and Robust Query Execution Plans62
LOGER, a learned optimizer towards generating efficient and robust plans, aiming at producing both efficient join orders and operators, is proposed, and can achieve a performance better than existing learned query optimizers.
- Uncertainty Quantification and Interpretability for Clinical Trial Approval Prediction57
This work devised a selective classification approach and integrated it with the Hierarchical Interaction Network, the state-of-the-art clinical trial prediction model, and demonstrated that incorporating uncertainty markedly enhances the model’s performance.
- OTOV2: Automatic, Generic, User-Friendly54
The second generation of Only-Train-Once (OTOv2), which first automatically trains and compresses a general DNN only once from scratch to produce a more compact model with competitive performance without fine-tuning, is proposed.
- 39
- Automatic Joint Structured Pruning and Quantization for Efficient Neural Network Training and Compression36
The framework GETA is presented, which automatically and efficiently performs joint structured pruning and quantization- aware training on any DNN, and introduces three key innovations: a quantization-aware dependency graph (QADG) that constructs a pruning search space for generic quantization-aware DNN, and a partially projected stochastic gradient method that guarantees layerwise bit constraints are satisfied.
- Uncertainty Quantification on Clinical Trial Outcome Prediction34
It is demonstrated that incorporating selective classification into clinical trial predictions markedly enhances the model's performance, as evidenced by significant upticks in pivotal metrics such as PR-AUC, F1, ROC-AUC, and overall accuracy, and is potentially setting a new benchmark in the field.
- Orthant Based Proximal Stochastic Gradient Method for $$\ell _1$$-Regularized Optimization30
On a large number of convex problems, OBProx-SG outperforms the existing methods comprehensively in the aspect of sparsity exploration and objective values, and the experiments on non-convex deep neural networks further demonstrate its superiority by achieving the solutions of much higher sparsity without sacrificing generalization accuracy.
- A Reduced-Space Algorithm for Minimizing $\ell_1$-Regularized Convex Functions30
A convergence guarantee is proved for the new method for minimizing the sum of a differentiable convex function and an $\ell_1$-norm regularizer and its efficiency is demonstrated on a large set of model prediction problems.
- Evolution of co-operation among mobile agents with different influence19
This work mainly focuses on the weighted influence on the mobility of agents in the prisoner’s dilemma game, and proves that co-operation level is promoted when the heterogeneous influence factor is considered.
- DRBench: A Realistic Benchmark for Enterprise Deep Research18
The DRBench benchmark is introduced, a benchmark for evaluating AI agents on complex, open-ended deep research tasks in enterprise settings that require identifying supporting facts from both the public web and private company knowledge base.
- Extensible Cross-Modal Hashing18
This work proposes a novel extensible cross-modal hashing (ECMH) model that has good forward compatibility, so there is no need to update old hash codes, and is extended to support new data categories using only new data by a well-designed “weak constraint incremental learning” algorithm.
- 17
- 17
- 17
- 16
- Single-image raindrop removal using concurrent channel-spatial attention and long-short skip connections16
A concurrent channel and spatial attention mechanism implemented by Squeeze & Excitation into the network is introduced to address the challenges of raindrop removal and the experimental results show that the proposed approach outperformed the existing ones.
- FaRSA for ℓ 1 -regularized convex optimization: local convergence and numerical experience15
An enhanced subproblem termination condition is introduced that allows us to prove that the iterates converge locally at a superlinear rate and the details of the publicly available C implementation are presented along with extensive numerical comparisons to other state-of-the-art solvers.
- 13
- 13
- 12
- A Generalized Alternating Method for Bilevel Learning under the Polyak-Łojasiewicz Condition11
A stationary metric is introduced for the considered bilevel problems, which generalizes the existing metric, and it is established that GALET achieves an (cid:15) -stationary metric for the considering problem within ˜ O ( ( cid: 15) − 1 ) iterations, which matches the iteration complexity of GD for smooth nonconvex problems.
- Single-image-based Rain Detection and Removal via CNN9
The experiments show that the designed CNN network outperforms state-of-the-art approaches on both synthesized and real-world images, which indicates the effectiveness of the proposed framework.
- Interactive design generation and optimization from generative adversarial networks in spatial computing8
A improved model for icon generation based on Auxiliary Classifier GANs performs well in presenting multiple feature representations of icons and helps the discriminative network recognize the core information of icon images more quickly.
- A Fast Reduced-Space Algorithmic Framework for Sparse Optimization8
The most popular type of regularization is `1-regularization in the last two decades, which has demonstrated its utility in improving the interpretation and accuracy of model estimation in both theoretical and practical aspects.
- OTOv3: Automatic Architecture-Agnostic Neural Network Training and Compression from Structured Pruning to Erasing Operators7
The third-generation Only-Train-Once (OTOv3) is introduced, which first automatically trains and compresses a general DNN through pruning and erasing operations, creating a compact and competitive sub-network without the need of fine-tuning.
- 6
- 6
- 5
- HESSO: Towards Automatic Efficient and User Friendly Any Neural Network Training and Pruning5
The numerical results showcase that HESSO can achieve competitive even superior performance to varying state-of-the-arts and support most DNN architectures and CRIC can effectively prevent the irreversible performance collapse and further enhance the performance of HESSO on certain applications.
- 5
- 4
- 4
- 4
- FSCNN: A Fast Sparse Convolution Neural Network Inference System3
This technical report study and present an efficient convolution neural network inference system to accelerate its forward pass by utilizing the fine-grained sparsity of compressed CNNs, and validate that FSCNN outperforms standard deep learning library PyTorch on popular CNN architectures such as VGG16 if sufficiently high sparsity exhibits.
- Efficient and Flexible Method for Reducing Moderate-Size Deep Neural Networks with Condensation2
A condensation reduction method to verify the feasibility of this idea in practical problems, thereby validating existing theories and reducing computational pressure and improving inference speed.
- S3Editor: A Sparse Semantic-Disentangled Self-Training Framework for Face Video Editing2
This work introduces S3Editor, a Sparse Semantic-disentangled Self-training framework for face video editing that adopts a self-training paradigm to enhance the training process through semi-supervision, and proposes a semantic disentangled architecture with a dynamic routing mechanism that accommodates diverse editing requirements.
- AdaContour: Adaptive Contour Descriptor with Hierarchical Representation2
Experiments show that AdaContour is able to represent shapes more accurately and robustly than other descriptors while retaining effectiveness, and is validated by integrating it into off-the-shelf detectors to enable instance segmentation.
- 2
- 3DHumanEdit: Multi-modal Body Part-aware Conditioning Information Integration for 3D Human Manipulation1
3DHumanEdit is proposed, a novel approach for 3D human body part-aware manipulation that outperforms existing methods in both editing fidelity and multi-view consistency, offering a robust solution for fine-grained 3D manipulation.
- 1
- 1
- 1
- 1
- An exposure guided background subtraction for smart camera1
Experimental results indicate the proposed model efficiently reduces false foreground pixels caused by auto-exposure, making it very preferable for smart camera without any extra hardware resources.
- DesInsert: Strategic descriptive term insertion fools text-to-image generation–
The Descriptive Term Insert is proposed along with two variants DesInsert-White and DesInsert-Black, designed for white-box and black-box settings, respectively, to introduce an optimized descriptive term into the input text, thereby preserving the semantics and enabling imperceptible attacks.
- RTNLW: Reversible and Tamper-Aware Natural Language Watermarking Scheme for Multi-Stage Transmission–
- –
- Design of an unmanned boat for waste collection and an automatic sorting device based on image recognition–
Validation experiments confirm that the system attains enhanced detection efficiency on embedded platforms, with stable multi-device collaboration latency maintained within 120 milliseconds, thereby providing a valuable technical reference for intelligent control system design in industrial inspection and smart agriculture.
- –
- –
- –
- –
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-14. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.