arXiv subjects

Dongjie Chen

Publications and source records attributed to Dongjie Chen.

7 recordsLinked to original sources

arXiv2026

Compiler-Grounded Hierarchical Diagnosis for LLM-Based Triton Kernel Optimization

Recent advances in large language models (LLMs) have enabled automated kernel generation and optimization, but most existing approaches rely on surface signals such as compilation feedback and profiling metrics. These signals reveal that a kernel is slow, but not why the backend compiler fails to realize a profitable optimization, especially on emerging accelerators such as NPUs. We therefore formulate kernel optimization as a progressive cross-layer diagnosis problem that links runtime symptoms to IR structure and compiler behavior before rewriting source. Based on this insight, we present our system, a compiler-grounded and hierarchical optimization framework for Triton kernels. the system escalates from lightweight pattern triage and profiling diagnosis to IR attribution and compiler-grounded analysis only when deeper evidence is needed, then proposes evidence-backed source-level rewrites. We implement the system on Triton for Ascend NPUs and evaluate it on 37 successfully converted entries from a standardized NPUKernelBench-derived Ascend 950 benchmark. Across these entries, the system attains a geometric-mean speedup of 4.35$\times$ and a median speedup of 2.73$\times$ from the initial to optimized Triton kernel; 22/37 exceed 2$\times$ and 13/37 exceed 5$\times$. The complete distribution ranges from near-baseline entries to large wins, motivating transparent reporting of the current system's scope and limitations.

cs.AI↗

arXiv2026

Empowering Source-Free Domain Adaptation via MLLM-Guided Reliability-Based Curriculum Learning

Existing SFDA methods struggle to fully use pre-trained knowledge and often rely on a single model's predictions or handcrafted prompts, limiting robustness under domain shift. Multimodal Large Language Models (MLLMs) offer a promising alternative: they encode rich visual-semantic knowledge and generalize well without task-specific tuning. However, their use in SFDA is hindered by instruction-following failures, inconsistent outputs, and high inference costs. We propose Reliability-based Curriculum Learning (RCL), a novel framework that distills robust supervision from multiple frozen MLLMs into a compact target model. RCL organizes adaptation as a three-stage curriculum that progressively incorporates pseudo-labels based on inter-model agreement and model confidence, enabling stable and noise-aware training. Our approach achieves state-of-the-art performance on standard SFDA datasets, Office-Home, DomainNet-126, and VisDA-C, outperforming zero-shot MLLMs, their ensembles, all without accessing source data or tuning foundation models. Our code is available at: https://github.com/Dong-Jie-Chen/RCL.

cs.LG↗

arXiv2025

MobilityGPT: Enhanced Human Mobility Modeling with a GPT model

Generative models have shown promising results in capturing human mobility characteristics and generating synthetic trajectories. However, it remains challenging to ensure that the generated geospatial mobility data is semantically realistic, including consistent location sequences, and reflects real-world characteristics, such as constraining on geospatial limits. We reformat human mobility modeling as an autoregressive generation task to address these issues, leveraging the Generative Pre-trained Transformer (GPT) architecture. To ensure its controllable generation to alleviate the above challenges, we propose a geospatially-aware generative model, MobilityGPT. We propose a gravity-based sampling method to train a transformer for semantic sequence similarity. Then, we constrained the training process via a road connectivity matrix that provides the connectivity of sequences in trajectory generation, thereby keeping generated trajectories in geospatial limits. Lastly, we proposed to construct a preference dataset for fine-tuning MobilityGPT via Reinforcement Learning from Trajectory Feedback (RLTF) mechanism, which minimizes the travel distance between training and the synthetically generated trajectories. Experiments on real-world datasets demonstrate MobilityGPT's superior performance over state-of-the-art methods in generating high-quality mobility trajectories that are closest to real data in terms of origin-destination similarity, trip length, travel radius, link, and gravity distributions. We release the source code and reference links to datasets at https://github.com/ammarhydr/MobilityGPT.

cs.LG↗

arXiv2024

YOLO-pdd: A Novel Multi-scale PCB Defect Detection Method Using Deep Representations with Sequential Images

With the rapid growth of the PCB manufacturing industry, there is an increasing demand for computer vision inspection to detect defects during production. Improving the accuracy and generalization of PCB defect detection models remains a significant challenge. This paper proposes a high-precision, robust, and real-time end-to-end method for PCB defect detection based on deep Convolutional Neural Networks (CNN). Traditional methods often suffer from low accuracy and limited applicability. We propose a novel approach combining YOLOv5 and multiscale modules for hierarchical residual-like connections. In PCB defect detection, noise can confuse the background and small targets. The YOLOv5 model provides a strong foundation with its real-time processing and accurate object detection capabilities. The multi-scale module extends traditional approaches by incorporating hierarchical residual-like connections within a single block, enabling multiscale feature extraction. This plug-and-play module significantly enhances performance by extracting features at multiple scales and levels, which are useful for identifying defects of varying sizes and complexities. Our multi-scale architecture integrates feature extraction, defect localization, and classification into a unified network. Experiments on a large-scale PCB dataset demonstrate significant improvements in precision, recall, and F1-score compared to existing methods. This work advances computer vision inspection for PCB defect detection, providing a reliable solution for high-precision, robust, real-time, and domain-adaptive defect detection in the PCB manufacturing industry.

cs.CV↗

arXiv2024

Assessment of different head tilt angles in volumetric modulated arc therapy for hippocampus-avoidance whole-brain radiotherapy

Purpose: We aims to examine the impact of different head tilt angles on the dose distribution in the whole-brain target area and organs at risk. It also aims to determine the head tilt angle to achieve optimal radiation therapy outcomes. Methods: CT images were collected from 8 brain metastases patients at 5 different groups of head tilt angle. The treatment plans were designed using the volumetric modulated arc therapy (VMAT) technique. The 5 groups of tilt angle were as follows: [0,10), [10,20), [20,30), [30,40), and [40,45]. The analysis involved assessing parameters such as the uniformity index, conformity index, average dose delivered to the target, dose coverage of the target, hot spots within the target area, maximum dose, and average dose received by organs at risk. Additionally, the study evaluated the correlation between hippocampal dose and other factors and established linear regression models. Results: Significant differences in dosimetric results were observed between the [40,45] and [0,10) head tilt angles. The [40,45] angle showed significant differences compared to the [0,10) angle in the average dose in the target area, dose uniformity, hotspots in the target area, maximum hippocampal dose, maximum dose in the lens, and average dose in the lens. There is a moderate correlation between the maximum dose in the hippocampi and the PTV length. Likewise, the mean dose in the hippocampi is significantly correlated with the hippocampi length. Conclusion: The VMAT plan with a head tilt angle of [40,45] met all dose constraints and demonstrated improved uniformity of the target area while reducing the dose to organs at risk. Furthermore, the linear regression models suggest that increasing the head tilt angle within the current range of [0,45] is likely to lead to a decrease in the average hippocampal dose.

physics.med-ph↗

arXiv2023

DPGOMI: Differentially Private Data Publishing with Gaussian Optimized Model Inversion

High-dimensional data are widely used in the era of deep learning with numerous applications. However, certain data which has sensitive information are not allowed to be shared without privacy protection. In this paper, we propose a novel differentially private data releasing method called Differentially Private Data Publishing with Gaussian Optimized Model Inversion (DPGOMI) to address this issue. Our approach involves mapping private data to the latent space using a public generator, followed by a lower-dimensional DP-GAN with better convergence properties. We evaluate the performance of DPGOMI on standard datasets CIFAR10 and SVHN. Our results show that DPGOMI outperforms the standard DP-GAN method in terms of Inception Score, Fréchet Inception Distance, and classification performance, while providing the same level of privacy. Our proposed approach offers a promising solution for protecting sensitive data in GAN training while maintaining high-quality results.

cs.LG↗

arXiv2022

Differentially Private Generative Adversarial Networks with Model Inversion

To protect sensitive data in training a Generative Adversarial Network (GAN), the standard approach is to use differentially private (DP) stochastic gradient descent method in which controlled noise is added to the gradients. The quality of the output synthetic samples can be adversely affected and the training of the network may not even converge in the presence of these noises. We propose Differentially Private Model Inversion (DPMI) method where the private data is first mapped to the latent space via a public generator, followed by a lower-dimensional DP-GAN with better convergent properties. Experimental results on standard datasets CIFAR10 and SVHN as well as on a facial landmark dataset for Autism screening show that our approach outperforms the standard DP-GAN method based on Inception Score, Fréchet Inception Distance, and classification accuracy under the same privacy guarantee.

cs.LG↗