Drift Happens: An Empirical Study of Neural Architecture Robustness to Temporal Distribution Shift
Department of Electrical Engineering and Computer Sciences, University of California, Berkeley（加州大学伯克利分校电气工程与计算机科学系）
Robin Holzinger, Riccardo Colletti
备注：
33 pages, 28 figures. Extended version of the QCDS 2026 proceedings paper; proceedings version to appear in Springer LNCS. Interactive website: https://drift-happens.org/
Real-world data distributions evolve over time, inducing temporal distribution shift that can substantially degrade the reliability of deployed machine learning systems. However, the extent to which architectural choices and their associated inductive biases affect temporal robustness remains insufficiently understood.
We present a systematic empirical comparison of temporal robustness across three heterogeneous, time-indexed domains encompassing image classification, multi-label text classification, and text regression tasks. Using a unified evaluation framework based on temporal drift matrices, we train models on cumulative historical data and evaluate their performance on both earlier and later time periods, thereby quantifying cross-temporal generalization. Our study spans model families ranging from simple multilayer perceptrons and convolutional networks to recurrent networks and pretrained Transformer-based encoders.
Collectively, the results show that architectural inductive biases systematically shape temporal robustness: models whose inductive biases lead them to exploit localized, highly discriminative features attain the highest in-distribution accuracy, yet those features are often the ones that change most over time, so these models degrade fastest, while pretrained encoders that draw on coarser, more stable representations drift more gradually. These observations offer practical guidance for selecting architectures for real-world systems subject to temporal drift.
Intel Corp.（英特尔公司）；Seoul National University（首尔国立大学）
While many real-world data streams imply that they change frequently in a nonstationary way, most of deep learning methods optimize neural networks on training data, and this leads to severe performance degradation when dataset shift happens. However, it is less possible to annotate or inspect newly streamed data by humans, and thus it is desired to measure model drift at inference time in an unsupervised manner. In this paper, we propose a novel method of model drift estimation by exploiting statistics of batch normalization layer on unlabeled test data. To remedy possible sampling error of s…
University of Wisconsin–Madison（威斯康星大学麦迪逊分校）；Microsoft Research（微软研究院）
In many real-world scenarios, we often deal with streaming data that is sequentially collected over time. Due to the non-stationary nature of the environment, the streaming data distribution may change in unpredictable ways, which is known as concept drift. To handle concept drift, previous methods first detect when/where the concept drift happens and then adapt models to fit the distribution of the latest data. However, there are still many cases that some underlying factors of environment evolution are predictable, making it possible to model the future concept drift trend of the streaming d…
Airbus Defence & Space Cybersecurity（空客防务与航天网络安全部门）
Recent advancements in Artificial Intelligence (AI) have brought new capabilities to behavioural analysis (UEBA) for cyber-security consisting in the detection of hostile action based on the unusual nature of events observed on the Information System.In our previous work (presented at C\&ESAR 2018 and FIC 2019), we have associated deep neural networks auto-encoders for anomaly detection and graph-based events correlation to address major limitations in UEBA systems. This resulted in reduced false positive and false negative rates, improved alert explainability, while maintaining real-time perf…
Microsoft Health and Life Sciences（微软健康与生命科学部）；Stanford Center for Artificial Intelligence in Medicine and Imaging（斯坦福医学与成像人工智能中心）
Clinical Artificial lntelligence (AI) applications are rapidly expanding worldwide, and have the potential to impact to all areas of medical practice. Medical imaging applications constitute a vast majority of approved clinical AI applications. Though healthcare systems are eager to adopt AI solutions a fundamental question remains: \textit{what happens after the AI model goes into production?} We use the CheXpert and PadChest public datasets to build and test a medical imaging AI drift monitoring workflow to track data and model drift without contemporaneous ground truth. We simulate drift in…
Safety goes first. Meeting and maintaining industry safety standards for robustness of artificial intelligence (AI) and machine learning (ML) models require continuous monitoring for faults and performance drops. Deep learning models are widely used in industrial applications, e.g., computer vision, but the susceptibility of their performance to environment changes (e.g., noise) \emph{after deployment} on the product, are now well-known. A major challenge is detecting data distribution shifts that happen, comparing the following: {\bf (i)} development stage of AI and ML models, i.e., train/val…
TriGuard: Testing Model Safety with Attribution Entropy, Verification, and Drift
Center for Data Science, New York University, New York, USA（数据科学中心，纽约大学，纽约，美国）；University of New Haven, Connecticut, USA（新罕布什尔大学，康涅狄格，美国）；Queens College, City University of New York, USA（皇后学院，纽约城市大学，美国）
Data Models for Dataset Drift Controls in Machine Learning With Optical Images
Fraunhofer HHI（弗劳恩霍夫HHI研究所）；Dotphoton AG（Dotphoton公司）；University of Glasgow（格拉斯哥大学）；HEPIA/HES-SO；Klinikum rechts der Isar（右侧-isar医院）；Helmholtz Zentrum Munich（慕尼黑亥姆霍兹中心）
A trained ML model is deployed on another `test' dataset where target feature values (labels) are unknown. Drift is distribution change between the training and deployment data, which is concerning if model performance changes. For a cat/dog image classifier, for instance, drift during deployment could be rabbit images (new class) or cat/dog images with changed characteristics (change in distribution). We wish to detect these changes but can't measure accuracy without deployment data labels. We instead detect drift indirectly by nonparametrically testing the distribution of model prediction co…
