ADATIME: A Benchmarking Suite for Domain Adaptation on Time Series Data

Abstract

Unsupervised domain adaptation methods aim to generalize well on unlabeled test data that may have a different (shifted) distribution from the training data. Such methods are typically developed on image data, and their application to time series data is less explored.Existing works on time series domain adaptation suffer from inconsistencies in evaluation schemes, datasets, and backbone neural network architectures. Moreover, labeled target data are usually employed for model selection, which violates the fundamental assumption of unsupervised domain adaptation. To address these issues, we develop a benchmarking evaluation suite (AdaTime) to systematically and fairly evaluate different domain adaptation methods on time series data. Specifically, we standardize the backbone neural network architectures and benchmarking datasets, while also exploring more realistic model selection approaches that can work with no labeled data or just few labeled samples. Our evaluation includes adapting state-of-the-art visual domain adaptation methods to time series data in addition to the recent methods specifically developed for time series data. We conduct extensive experiments to evaluate 10 state-of-the-art methods on four representative datasets spanning 20 cross-domain scenarios. Our results suggest that with careful selection of hyper-parameters, visual domain adaptation methods are competitive with methods proposed for time series domain adaptation. In addition, we find that hyper-parameters could be selected based on realistic model selection approaches. Our work unveils practical insights for applying domain adaptation methods on time series data and builds a solid foundation for future works in the field. The code is available at github.com/emadeldeen24/AdaTime.

High-performance zero-standby-power-consumption-under-bending pressure sensors for artificial reflex arc

“…We also consider four unsupervised DA methods for time series: CODATS (Wilson et al, 2020), adversarial spectral kernel matching for unsupervised time series domain adaptation (AdvSKM) (Liu and Xue, 2021b), and CLUDA (Ozyurt et al, 2022). We additionally consider source-domainonly training (no transfer) implemented by (Ragab et al, 2022). Baselines for uniDA.…”

“…We run test on a NVIDIA GeForce RTX 3090 graphic card. We implement the time feature extracto via a convolutional network (CNN) (Ragab et al, 2022). This configuration remains the same across all methods so that the difference in prediction performance is attributed to algorithm.…”

Self-supervised Learning for Label-Efficient Sleep Stage Classification: A Comprehensive Evaluation

“…This distribution shift can be caused by a different data collection methodology or differences in subjects' health status. To deal with this challenging scenario, some recent works proposed transfer learning and unsupervised domain adaptation algorithms to mitigate the domain shift [38]- [41].…”

SEnsor Alignment for Multivariate Time-Series Unsupervised Domain Adaptation

“…VRADA (Purushotham et al 2017) learned temporal features via a variational RNN and reduced the discrepancy by adversarial based methods. ADATIME (Ragab et al 2022b) evaluated various CNN models to capture temporal dynamics. However, these works were designed specifically for the data from one source, i.e., the data from same distributions, which are inapplicable for the scenarios requiring multiple sensors.…”

“…Different from UTS data, MTS data originate from multiple sensors where data from different sensors follow various data distributions. By treating all sensors as a whole, existing UDA methods can consider the global distributions of sensors (Ragab et al 2022b). But they cannot take the sensor-level distributions into account, leading to the misalignment of each sensor.…”