Pan Hui | TU Delft Repository

SMOOTH

Scalable Multitask Offloading with Backbone Sharing

Conference paper (2026) - Wei Geng, Xiang Su, Nitinder Mohan, Jorg Ott, Pan Hui

Intelligent mobile applications are often constrained by limited on-device hardware and by the latency and bandwidth overhead of cloud offloading. Offloading computation-intensive deep learning tasks to edge servers can potentially mitigate these challenges. However, existing systems struggle to scale as the numbers of tasks and users grow. This limitation stems from a fundamental conflict: requests in edge settings are typically sparse, heterogeneous, and demand immediate processing to minimize latency. However, GPUs operate most efficiently on dense, homogeneous batches, a condition that edge traffic rarely provides. Most existing approaches, including cloud-oriented solutions like Multi-Instance GPU, fail to resolve this tension because edge devices typically lack the advanced features available in high-end cloud GPUs. This paper contributes SMOOTH, a scalable multitask offloading system that introduces a Sparsity-to-Density Abstraction to resolve this conflict. By decomposing deep learning models into a shared backbone and lightweight task-specific heads, SMOOTH transforms sparse, heterogeneous request streams into dense, homogeneous computation blocks amenable to efficient batching. This allows highly efficient cross-task batching without prohibitive accumulation delays. To further optimize responsiveness, SMOOTH employs a compute-lightweight (O(1)), sparsity-aware adaptive scheduler that dynamically balances inference throughput and end-to-end latency based on queue dynamics. Our evaluations demonstrate up to 2.21× higher throughput, 45% lower memory usage, and up to 82% reduced latency across diverse arrival patterns compared to baselines. ...

DeepPick

A Deep Learning Approach to Unveil Outstanding Users Ranking with Public Attainable Features

Journal article (2023) - Wanda Li, Zhiwei Xu, Yi Sun, Qingyuan Gong, Yang Chen, Aaron Yi Ding, Xin Wang, Pan Hui

Outstanding users (OUs) denote the influential, 'core' or 'bridge' users in online social networks. How to accurately detect and rank them is an important problem for third-party online service providers and researchers. Conventional efforts, ranging from early graph-based algorithms to recent machine learning-based approaches, typically rely on an entire social network's information. However, for privacy-conscious users or newly-registered users, such information is not easily accessible. To address this issue, we present DeepPick, a novel framework that considers both the generalization and specialization in the detection task of OUs. For generalization, we introduce deep neural networks to capture dynamic features of the users. For specialization, we leverage the traditional descriptive features to make use of public information about users. Extensive experiments based on real-world datasets demonstrate that our approach achieves a high efficacy of detection performance against the state-of-the-art. ...

Metaverse for Connected and Automated Vehicles and Intelligent Transportation Systems [From the Guest Editors]

Review (2023) - Pengyuan Zhou, Lik Hang Lee, Zhi Liu, Hang Qiu, Tristan Braud, Aaron Yi Ding, Sasu Tarkoma, Pan Hui

The metaverse aims to blur the boundary between the physical world and digital content. To achieve this goal, the metaverse relies heavily on extended reality (XR), the Internet of Things, and communication technologies. Concurrently, connected vehicles and intelligent transportation systems (ITSs) are envisioned as the future paradigm of driving and becoming reality thanks to increasingly powerful onboard vehicular processing capacity and advanced vehicle-to-everything networking technologies. ...

Understanding the Working Time of Developers in IT Companies in China and the United States

Journal article (2020) - Jiayun Zhang, Yang Chen, Qingyuan Gong, Aaron Yi Ding, Yu Xiao, Xin Wang, Pan Hui

We identified three temporal patterns shown in commit activities among Chinese and American companies and found that Chinese businesses are more likely to follow long work hours than American ones. We also conducted a survey on the trends of, reasons for, and results of overtime work. Our study could provide references for developers to choose workplaces and for companies to make regulations. ...

Enhancing the internet of things with knowledge-driven software-defined networking technology

Future perspectives

Review (2020) - Yuhong Li, Xiang Su, Aaron Yi Ding, Anders Lindgren, Xiaoli Liu, Christian Prehofer, Jukka Riekki, Rahim Rahmani, Sasu Tarkoma, Pan Hui

The Internet of Things (IoT) connects smart devices to enable various intelligent services. The deployment of IoT encounters several challenges, such as difficulties in controlling and managing IoT applications and networks, problems in programming existing IoT devices, long service provisioning time, underused resources, as well as complexity, isolation and scalability, among others. One fundamental concern is that current IoT networks lack flexibility and intelligence. A network-wide flexible control and management are missing in IoT networks. In addition, huge numbers of devices and large amounts of data are involved in IoT, but none of them have been tuned for supporting network management and control. In this paper, we argue that Software-defined Networking (SDN) together with the data generated by IoT applications can enhance the control and management of IoT in terms of flexibility and intelligence. We present a review for the evolution of SDN and IoT and analyze the benefits and challenges brought by the integration of SDN and IoT with the help of IoT data. We discuss the perspectives of knowledge-driven SDN for IoT through a new IoT architecture and illustrate how to realize Industry IoT by using the architecture. We also highlight the challenges and future research works toward realizing IoT with the knowledge-driven SDN. ...