Graph-OPU: An FPGA-Based Overlay Processor for Graph Neural Networks

Chen, Ruiqi; Zhang, Haoyang; Ma, Yuhanxiao; Tang, Enhao; Li, Shun; Zhu, Yanxiang; Yu, Jun; Wang, Kun

doi:10.1145/3543622.3573152

Cited by 5 publications

(3 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Hence, many efforts have been made to develop fair GNNs. According to the stage at which the debiasing process occurs, the existing methods could be split into the pre-processing, in-processing, and post-processing methods [3]. Pre-processing methods remove bias before GNN training occurs by targeting the input graph structure, input features, or both.…”

Section: Fairness Of Graphmentioning

confidence: 99%

“…Debiasing for a specific task in the pre-training phase is inflexible, and maintaining a specific PGM for each task is inefficient. Besides, most existing fairness methods lack theoretical analysis and guarantees [3,20], meaning that they do not provide a practical guarantee, i.e., provable lower bounds on the fairness of model prediction. This is significant for determining whether to deploy models in practical scenarios [5,19,34,35].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Endowing Pre-trained Graph Models with Provable Fairness

Zhang,

et al. 2024

Proceedings of the ACM Web Conference 2024

View full text Add to dashboard Cite

Pre-trained graph models (PGMs) aim to capture transferable inherent structural properties and apply them to different downstream tasks. Similar to pre-trained language models, PGMs also inherit biases from human society, resulting in discriminatory behavior in downstream applications. The debiasing process of existing fair methods is generally coupled with parameter optimization of GNNs. However, different downstream tasks may be associated with different sensitive attributes in reality, directly employing existing methods to improve the fairness of PGMs is inflexible and inefficient. Moreover, most of them lack a theoretical guarantee, i.e., provable lower bounds on the fairness of model predictions, which directly provides assurance in a practical scenario. To overcome these limitations, we propose a novel adapter-tuning framework that endows pre-trained Graph models with Provable fAiRness (called GraphPAR 1 ). GraphPAR freezes the parameters of PGMs and trains a parameter-efficient adapter to flexibly improve the fairness of PGMs in downstream tasks. Specifically, we design a sensitive semantic augmenter on node representations, to extend the node representations with different sensitive attribute semantics for each node. The extended representations will be used to further train an adapter, to prevent the propagation of sensitive attribute semantics from PGMs to task predictions. Furthermore, with GraphPAR, we quantify whether the fairness of each node is provable, i.e., predictions are always fair within a certain range of sensitive attribute semantics. Experimental evaluations on real-world datasets demonstrate that GraphPAR achieves state-of-the-art prediction performance and fairness on node classification task. Furthermore, based on our GraphPAR, around 90% nodes have provable fairness.

show abstract

Section: Fairness Of Graphmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Endowing Pre-trained Graph Models with Provable Fairness

Zhang,

et al. 2024

Proceedings of the ACM Web Conference 2024

View full text Add to dashboard Cite

show abstract

“…Quantization is an effective approach to overcome the challenges posed by graph data, not only for embedded devices but also for GPU-based applications. Tango [76] aims to accelerate GNN training on GPU systems using techniques such as GEMM, SPMM, and SDDMM. The authors propose a set of rules and stochastic rounding for GPUs to speed up training without compromising accuracy.…”

Section: Quantization Approaches For Gnnsmentioning

confidence: 99%

A Survey of Computationally Efficient Graph Neural Networks for Reconfigurable Systems

Kose,

Nunez-Yanez,

Piechocki

et al. 2024

Preprint

View full text Add to dashboard Cite

Graph Neural Networks (GNNs) are powerful models capable of managing intricate connections in non-Euclidean data, such as social networks, physical systems, chemical structures, and communication networks. Despite their effectiveness, the large-scale and complex nature of graph data demands substantial computational resources and high performance during both training and inference stages, presenting significant challenges, particularly in the context of embedded systems. Recent studies on GNNs have investigated both software and hardware solutions to enhance computational efficiency. Earlier studies on Deep Neural Networks (DNNs) have indicated that methods like reconfigurable hardware and quantization are beneficial in addressing these issues. Unlike DNN research, studies on efficient computational methods for GNNs are less developed and require more exploration. This survey reviews the latest developments in quantization and FPGA-based acceleration for GNNs, showcasing the capabilities of reconfigurable systems (often FPGAs) to offer customized solutions in environments marked by significant sparsity and the necessity for dynamic load management. It also emphasizes the role of quantization in reducing both computational and memory demands through the use of fixed-point arithmetic and streamlined vector formats. This paper concentrates on low-power, resource-limited devices over general hardware accelerators and reviews research applicable to embedded systems. Additionally, it provides a detailed discussion of potential research gaps, foundational knowledge, obstacles, and prospective future directions.

show abstract