Online Active Learning for Drifting Data Streams

CADM+: Confusion-Based Learning Framework With Drift Detection and Adaptation for Real-Time Safety Assessment

Real-time safety assessment (RTSA) of dynamic systems holds substantial implications across diverse fields, including industrial and electronic applications. However, the complexity and rapid flow nature of data streams, coupled with the expensive label cost and pose significant challenges. To address these issues, a novel confusion-based learning framework, termed confusion-and-detection method plus (CADM+), is proposed in this article. When drift occurs, the model is updated with uncertain samples, which may cause confusion between existing and new concepts, resulting in performance differences. The cosine similarity is used to measure the degree of such conceptual confusion in the model. Furthermore, the change of standard deviation within a fixed-size cosine similarity window is introduced as an indicator for drift detection. Theoretical demonstrations show the asymptotic increase of cosine similarity. In addition, the approximate independence of the change in standard deviation with the number of trained samples is indicated. Finally, the extreme value theory (EVT) is applied to determine the threshold of judging drifts. Several experiments are conducted to verify its effectiveness. Experimental results prove that the proposed framework is more suitable for RTSA tasks compared with state-of-the-art algorithms. The source code is available at https://github.com/THUFDD/CADM-plus.

A Survey on Scenario Theory, Complexity, and Compression-Based Learning and Generalization

This work investigates formal generalization error bounds that apply to support vector machines (SVMs) in realizable and agnostic learning problems. We focus on recently observed parallels between probably approximately correct (PAC)-learning bounds, such as compression and complexity-based bounds, and novel error guarantees derived within scenario theory. Scenario theory provides nonasymptotic and distributional-free error bounds for models trained by solving data-driven decision-making problems. Relevant theorems and assumptions are reviewed and discussed. We propose a numerical comparison of the tightness and effectiveness of theoretical error bounds for support vector classifiers trained on several randomized experiments from 13 real-life problems. This analysis allows for a fair comparison of different approaches from both conceptual and experimental standpoints. Based on the numerical results, we argue that the error guarantees derived from scenario theory are often tighter for realizable problems and always yield informative results, i.e., probability bounds tighter than a vacuous [0, 1] interval. This work promotes scenario theory as an alternative tool for model selection, structural-risk minimization, and generalization error analysis of SVMs. In this way, we hope to bring the communities of scenario and statistical learning theory closer, so that they can benefit from each other's insights.

Codesign of FDI Attacks Detection, Isolation, and Mitigation for Complex Microgrid Systems: An HBF-NN-Based Approach

The primary purpose of this article is to design an intelligent false data injection (FDI) attacks detection, isolation, and mitigation scheme for a class of complex microgrid systems with electric vehicles (EVs). First, a networked microgrid with an EV model is well established, which takes load disturbance, wind generation fluctuation, and FDI attacks into account so as to truly reflect the operation process of the complex system. Then, an intelligent hyper basis function neural network (HBF-NN) observer is designed to accurately estimate the state of the microgrids, learn, and reconstruct the possible attack signal online. Subsequently, a novel HBF-NN-based H∞ controller is skillfully designed to mitigate the negative impact of FDI attacks online, so as to ensure the normal operation of the complex systems in an unreliable network environment. Finally, a two-stage integrated intelligent detection and maintenance algorithm is summarized and one simulation is presented to provide tangible evidence of the feasibility and superiority of the proposed FDI attacks detection, isolation, and mitigation methodology.

Dynamic Submodular-Based Learning Strategy in Imbalanced Drifting Streams for Real-Time Safety Assessment in Nonstationary Environments

The design of real-time safety assessment (RTSA) approaches in nonstationary environments is meaningful to reduce the possibility of significant losses. However, several challenging problems are needed to be well considered. The performance of existing approaches will be negatively affected in the settings of imbalanced drifting streams. In this case, the model design with the incremental update should also be explored. Furthermore, the query strategy should also be well-designed. This article investigates a dynamic submodular-based learning strategy to address such issues. Specifically, an efficient incremental update procedure is designed with the structure of the broad learning system (BLS), which is beneficial to the detection of concept drift. Furthermore, a novel dynamic submodular-based annotation with an activation interval strategy is proposed to select valuable samples in imbalanced drifting streams. The lower bound of annotation value is also proven theoretically with a novel drift adaption mechanism. Numerous experiments are conducted with the realistic data of JiaoLong deep-sea manned submersible. The experimental results show that the proposed approach can achieve better assessment accuracy than typical existing approaches.