DOMINO: Domain-Invariant Hyperdimensional Classification for Multi-Sensor Time Series Data
DOMINO: Domain-Invariant Hyperdimensional Classification for Multi-Sensor Time Series Data
复制标题
DOI：
10.1109/iccad57390.2023.10323848
复制
发表时间：
2023-08
期刊：
影响因子：
--
通讯作者：
Junyao Wang;Luke Chen;M. A. Faruque
中科院分区：
文献类型：
--
作者：
Junyao Wang;Luke Chen;M. A. Faruque
随着物联网的快速发展，许多现实世界的应用程序利用了异质连接的传感器来捕获时间序列信息。基于边缘的机器学习（ML）方法通常用于分析本地收集的数据。但是，跨数据驱动的ML方法的基本问题是分配变化。当将模型部署在数据分布上与训练的数据分布不同，并且可能会降低模型性能时，就会发生这种情况。此外，已经提出了越来越复杂的深度神经网络（DNN），以捕获多传感器时间序列数据中的空间和时间依赖性，需要超出当今边缘设备能力的密集计算资源。尽管已引入了受脑启发的高度计算（HDC）作为基于边缘学习的轻量级解决方案，但现有的HDC也容易受到分配转移挑战的影响。在本文中，我们提出了Domino，这是一个新型的HDC学习框架，该框架解决了嘈杂的多传感器时间序列数据中的分布转移问题。多米诺在高维空间上利用高效和并行矩阵操作，以动态识别和滤除域变化的尺寸。我们对各种多传感器时间序列分类任务的评估表明，多米诺骨牌的准确性平均比最先进的（SOTA）基于DNN的域概括技术高2.04％，并且提供了$ 16.34 \ times $ $ $ $ $ $ $ $ $和$ 2.89 \ times $ $更快的推理。更重要的是，在从部分标记的数据和高度不平衡的数据中学习时，多米诺骨牌的性能明显更好，并且比SOTA DNNS提供了$ 10.93 \ times $稳健性。
With the rapid evolution of the Internet of Things, many real-world applications utilize heterogeneously connected sensors to capture time-series information. Edge-based machine learning (ML) methodologies are often employed to analyze locally collected data. However, a fundamental issue across data-driven ML approaches is distribution shift. It occurs when a model is deployed on a data distribution different from what it was trained on, and can substantially degrade model performance. Additionally, increasingly sophisticated deep neural networks (DNNs) have been proposed to capture spatial and temporal dependencies in multi-sensor time series data, requiring intensive computational resources beyond the capacity of today's edge devices. While brain-inspired hyperdimensional computing (HDC) has been introduced as a lightweight solution for edge-based learning, existing HDCs are also vulnerable to the distribution shift challenge. In this paper, we propose DOMINO, a novel HDC learning framework addressing the distribution shift problem in noisy multi-sensor time-series data. DOMINO leverages efficient and parallel matrix operations on high-dimensional space to dynamically identify and filter out domain-variant dimensions. Our evaluation on a wide range of multi-sensor time series classification tasks shows that DOMINO achieves on average 2.04% higher accuracy than state-of-the-art (SOTA) DNN-based domain generalization techniques, and delivers $16.34\times$ faster training and $2.89\times$ faster inference. More importantly, DOMINO exhibits notably better performance when learning from partially labeled data and highly imbalanced data, and provides $10.93\times$ higher robustness against hardware noises than SOTA DNNs.