JGJiawei Guan

Jiawei Guan

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Jiawei Guan has a total of 29 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    CEDAR: Compressed-Object Parallel Retrieval for Scalable AI Systems

    by Jiawei Guan, Feng Zhang, Yancong Wang, Bingsheng He, Jidong Zhai, Xiaoyong Du

  • Open Access

    Parallel-Aware Early-Stopping Metrics for Hyperparameter Optimization

    by Jiawei Guan

    Parallel and distributed HPO systems evaluate many configurations concurrently, but their throughput depends heavily on when a scheduler stops weak trials and reallocates scarce CPUs, GPUs, and memory. Existing systems commonly rank live trials by validation loss, treating the early-stop metric as an implementation detail. This demo reframes metric selection as a parallel resource-scheduling problem. We present an interactive workbench for inspecting how training loss, validation loss, uncertainty-aware metrics, and stage-adaptive metrics affect Successive Halving and Hyperband decisions under different worker budgets. Across NAS-Bench-201, LCBench, and HPOBench, the underlying study shows that training loss can improve early-stage HPO outcomes by up to 24.76% over validation loss, and uncertainty-aware metrics can further improve outcomes by up to 4% under constrained budgets. The demo helps parallel-computing researchers reason about scheduler reliability, wasted worker time, and budget allocation policies for large-scale model tuning.

  • Open Access

    L3: A GPU-Native Co-Designed Data Format for Learned Lossless Lightweight Compression

    by Youyang Xia, Feng Zhang, Junda Pan, Yihao Liu, Jiawei Guan, Huanchen Zhang, Xiaoyong Du

    Learned Compression achieves strong CPU performance but lacks a GPU-native format, limiting its use in GPU analytics. We present L3, a GPU-native Learned Lossless Lightweight Compression format that enables end-to-end on-device processing with efficient compression, high-throughput decompression, and fast random access on GPU. On NVIDIA GPUs, a warp is a group of 32 threads; we refer to each thread as a lane (lane id 0–31), and call a layout lane-major when each lane's words are stored contiguously. L3 introduces three tightly coupled components built around the SLAP Vertical layout. First, the L3 Storage Layout (SLAP) stores bit-packed residual streams in a lane-major organization, i.e., residual words are laid out lane by lane so each warp lane consumes a contiguous word sequence in memory, exploiting the GPU L1 sector cache for implicit prefetching and high reuse during unpacking. Second, the Warp-Cooperative Learned Decompression Module maps each partition to one thread block and decodes warp tiles using per-lane bit readers, branchless bit extraction, and a bit-exact no-FMA FP64 finite-difference predictor. Third, the GPU-Native Learned Compression Pipeline builds adaptive partitions via bulk delta-bits analysis, scan/compaction, and an odd-even GPU merge loop, then packs residuals directly into the final SLAP Vertical layout on the device. L3 achieves high performance on modern GPUs. It encodes 3–6× faster than Tile and FastLanes-GPU and sustains 1.08–1.90 TB/s decompression throughput, comparable to the fastest lightweight GPU codecs. On correlated datasets, L3 reaches up to 77× compression while remaining competitive on weakly correlated inputs. For random access, L3 maintains 1.2–2.6 Billion queries/s and outperforms Tile-DFOR/Tile-RFOR by 5–10×. On SSB with unified query plans, L3 achieves the lowest average latency (1.14 ms), matching or outperforming state-of-the-art GPU baselines.

  • Open Access

    Enabling Efficient Update on Rule-Based Compressed Graph

    by Lin Feng, Feng Zhang, Zheng Chen, Yuxin Tang, Jiawei Guan, Xiaowei Zhu, Xiaoyong Du

    Dynamic graphs, pivotal in applications ranging from social networks to biological systems, pose significant challenges in storage and update efficiency due to their continuous evolution through insertions and deletions. Traditional graph compression methods, primarily optimized for static graphs, often struggle in dynamic environments, leading to costly decompression-recompression cycles during updates. To address this limitation, we propose a novel theoretical framework that enables efficient direct updates on rule-based compressed graphs. Additionally, we introduce a dynamic graph processing framework that balances space efficiency, update responsiveness, and query performance. Our framework treats updates as lightweight, localized modifications, sustains compression via background cleanup, and preserves query performance by ensuring structural integrity. Our method achieves significant memory savings, reducing usage by an average of 50.1% over state-of-the-art dynamic compressed graph systems, while demonstrating highly competitive update throughput and query performance. These advancements establish a scalable, high-performance solution for dynamic graph management, effectively bridging the gap between space efficiency, updates responsiveness and query performance.


  • Top Cited Papers

    Breaking the Edge: Enabling Efficient Neural Network Inference on Integrated Edge Devices

    by Feng Zhang, Chenyang Zhang, Jiawei Guan, Qiangjun Zhou, Kuangyu Chen, Xiao Zhang, Bingsheng He, Jidong Zhai, Xiaoyong Du

    Edge computing has gained widespread attention in cloud computing due to the increasing demands of AIoT applications and the evolution of edge architectures. One prevalent application in this domain is neural network inference on edge for computing and processing. This article presents an in-depth exploration of inference on integrated edge devices and introduces EdgeNN, a groundbreaking solution for inference specifically designed for CPU-GPU integrated edge devices. EdgeNN offers three key innovations. First, EdgeNN adaptively employs zero-copy optimization by harnessing unified physical memory. Second, EdgeNN introduces an innovative approach to CPU-GPU hybrid execution tailored for inference tasks. This technique enables concurrent CPU and GPU operation, effectively leveraging edge platforms’ computational capabilities. Third, EdgeNN adopts a finely tuned adaptive inference tuning technique that analyzes complex inference structures. It divides computations into sub-tasks, intelligently assigning them to the two processors for better performance. Experimental results demonstrate EdgeNN's superiority across six popular neural network inference processing. EdgeNN delivers average speed improvements of 3.97×, 4.10×, 3.12×, and 8.80× when compared to inference on four distinct edge CPUs. Furthermore, EdgeNN achieves significant time advantages compared to the direct execution of original programs. This improvement is attributed to better unified memory utilization (44.37%) and the innovative CPU-GPU hybrid execution approach (17.91%). Additionally, EdgeNN exhibits superior energy efficiency, providing 29.14× higher energy efficiency than edge CPUs and 5.70× higher energy efficiency than discrete GPUs. EdgeNN is now open source at https://github.com/ChenyangZhang-cs/EdgeNN.


  • Enabling Efficient Deep Learning on MCU With Transient Redundancy Elimination

    by Jiesong Liu, Feng Zhang, Jiawei Guan, Hsin-Hsuan Sung, Xiaoguang Guo, Saiqin Long, Xiaoyong Du, Xipeng Shen

    Deploying deep neural networks (DNNs) with satisfactory performance in resource-constrained environments is challenging. This is especially true of microcontrollers due to their tight space and computational capabilities. However, there is a growing demand for DNNs on microcontrollers, as executing large DNNs on microcontrollers is critical to reducing energy consumption, increasing performance efficiency, and eliminating privacy concerns. This paper presents a novel and systematic data redundancy elimination method to implement efficient DNNs on microcontrollers through innovations in computation and space optimization. By making the optimization itself a trainable component in the target neural networks, this method maximizes performance benefits while keeping the DNN accuracy stable. Experiments are performed on two microcontroller boards with three popular DNNs, namely CifarNet, ZfNet and SqueezeNet. Experiments show that this solution eliminates more than 96% of computations in DNNs and makes them fit well on microcontrollers, yielding 3.4-5× speedup with little loss of accuracy.


  • Open Access

    Homomorphic Compression: Making Text Processing on Compression Unlimited

    by Jiawei Guan, Feng Zhang, Siqi Ma, Kuangyu Chen, Yihua Hu, Yuxing Chen, Anqun Pan, Xiaoyong Du

    Lossless data compression is an effective way to handle the huge transmission and storage overhead of massive text data. Its utility is even more significant today when data volumes are skyrocketing. The concept of operating on compressed data infuses new blood into efficient text management by enabling mainly access-oriented text processing tasks to be done directly on compressed data without decompression. Facing limitations of the existing compressed text processing schemes such as limited types of operations supported, low efficiency, and high space occupation, we address these problems by proposing a homomorphic compression theory. It enables the generalization and characterization of algorithms with compression processing capabilities. On this basis, we develop HOCO, an efficient text data management engine that supports a variety of processing tasks on compressed text. We select three representative compression schemes and implement them combined with homomorphism in HOCO. HOCO supports the extension of homomorphic compression schemes through a modular and object-oriented design and has convenient interfaces for text processing tasks. We evaluate HOCO on six real-world datasets. The three schemes implemented in HOCO show trade-offs in terms of compression ratio, supported operation types, and efficiency. Experiments also show that HOCO can achieve higher throughput in random access and modification operations (averagely 9.18× than the state-of-the-art) and lower latency in text analytic tasks (averagely 7.16× than processing on uncompressed text) without compromising compression efficacy.


  • Open Access

    CompressGraph: Efficient Parallel Graph Analytics with Rule-Based Compression

    by Zheng Chen, Feng Zhang, JiaWei Guan, Jidong Zhai, Xipeng Shen, Huanchen Zhang, Wentong Shu, Xiaoyong Du

    Modern graphs exert colossal time and space pressure on graph analytics applications. In 2022, Facebook social graph reaches 2.91 billion users with trillions of edges. Many compression algorithms have been developed to support direct processing on compressed graphs to address this challenge. However, previous graph compression algorithms do not focus on leveraging redundancy in repeated neighbor sequences, so they do not save the amount of computation for graph analytics. We develop CompressGraph, an efficient rule-based graph analytics engine that leverages data redundancy in graphs to achieve both performance boost and space reduction for common graph applications. CompressGraph has three advantages over previous works. First, the rule-based abstraction of CompressGraph supports the reuse of intermediate results during graph traversal, thus saving time. Second, CompressGraph has intense expressiveness to support a wide range of graph applications. Third, CompressGraph scales well under high parallelism because the context-free rules have few dependencies. Experiments show that CompressGraph provides significant performance and space benefits on both CPUs and GPUs. On evaluating six typical graph applications, CompressGraph can achieve 1.97× speedup on the CPU, while 3.95× speedup on the GPU, compared to the state-of-the-art CPU and GPU methods, respectively. Moreover, CompressGraph can save an average of 71.27% memory savings on CPU and 70.36 on GPU.


  • Expanding the Edge: Enabling Efficient Winograd CNN Inference With Deep Reuse on Edge Device

    by Feng Zhang, Ruofan Wu, Jiawei Guan, Zhen Zheng, Xiaoguang Guo, Xiao Zhang, Xiaoyong Du, Xipeng Shen

    Deep learning on edge devices is becoming increasingly important, especially with the explosion of IoT devices. For example, the total number of devices connected to IoT reaches 29 billion in 2022. Convolutional neural networks (CNNs), as common deep learning representatives, are among the most popular neural networks in knowledge and data engineering. However, CNN employs a high degree of computing. In comparison to the training phase, the inference process is more frequently done on low-power computing equipments, such as edge devices. The limited computing resource and high computation pressure limit the effective use of CNN algorithms at the edge. Fortunately, a minimal filtering algorithm called Winograd can reduce convolution calculations by minimizing multiplication operations. We find that Winograd convolution can be accelerated further by deep reuse technique, which reuses the similar data and computation processes. In this paper, we propose a new inference method, called DREW, which combines deep reuse with Winograd for further accelerating CNNs. DREW handles three difficulties. First, it can detect the similarities from the complex minimal filtering patterns by clustering. Second, it reduces the online clustering cost in a reasonable range. Third, it provides an adjustable method in clustering granularity balancing the performance and accuracy. We perform evaluation on Raspberry PI and NVIDIA Jetson AGX Xavier edge devices, and experiments show that on five popular networks, 1) DREW further accelerates the Winograd convolution by an average of 8.27× speedup. Even for the highly parallel Winograd implementation, DREW still can provide 2.21× speedup. 2) When DREW is applied to end-to-end Winograd CNN inferences, DREW achieves 5.94× the average performance speedup with no ( $< $ 0.4%) accuracy loss. 3) Energy consumption is an important factor for inference in practice. DREW reduces the number of convolution operations to 10% of the original operations, thus achieving up to 60% energy-efficiency benefits than the original Winograd inference.