arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2407.10016cs.CVcs.AI
收藏

表征视觉任务中边缘模型与高精度基准模型之间的差异

Characterizing Disparity Between Edge Models and High-Accuracy Base Models for Vision Tasks

University of North Carolina at Chapel Hill(北卡罗来纳大学教堂山分校)

Zhenyu Wang, Shahriar Nirjon

AI总结:

针对边缘模型与高精度基准模型的差异解释需求,研究提出含DELTA网络的XDELTA可解释工具,经大规模数据与实际部署验证,可有效解释两类模型的差异。

摘要:

边缘设备的计算能力差异很大,可支持多种多样的边缘AI模型,这也引出了一个问题:针对同一任务,边缘模型与高精度(基准)模型有何不同?我们推出了XDELTA,这是一款新型可解释AI工具,用于解释高精度基准模型与计算高效但精度较低的边缘模型之间的差异。为此,我们提出了一种基于学习的方法来表征模型差异,即DELTA网络(差异网络),它以紧凑的形式补充边缘网络的特征表征能力。为构建DELTA,我们提出了一种稀疏优化框架,提取基准模型的核心信息,以确保DELTA的紧凑性与充足的特征表征能力;同时实现了一种负相关学习方法,确保其能与边缘模型形成互补。我们开展了综合评估以测试XDELTA解释模型差异的能力,评估使用了超过120万张图像、24个模型,并通过6名参与者测试了实际部署效果。XDELTA可通过几何分析与概念层面分析,出色地解释基准模型与边缘模型(任意模型对以及压缩后的基准模型)之间的差异,在实际应用中被证明有效。

英文摘要:

Edge devices, with their widely varying capabilities, support a diverse range of edge AI models. This raises the question: how does an edge model differ from a high-accuracy (base) model for the same task? We introduce XDELTA, a novel explainable AI tool that explains differences between a high-accuracy base model and a computationally efficient but lower-accuracy edge model. To achieve this, we propose a learning-based approach to characterize the model difference, named the DELTA network, which complements the feature representation capability of the edge network in a compact form. To construct DELTA, we propose a sparsity optimization framework that extracts the essence of the base model to ensure compactness and sufficient feature representation capability of DELTA, and implement a negative correlation learning approach to ensure it complements the edge model. We conduct a comprehensive evaluation to test XDELTA's ability to explain model discrepancies, using over 1.2 million images and 24 models, and assessing real-world deployments with six participants. XDELTA excels in explaining differences between base and edge models (arbitrary pairs as well as compressed base models) through geometric and concept-level analysis, proving effective in real-world applications.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑