Adaptive learning under concept drift for multiple data streams
- Publication Type:
- Thesis
- Issue Date:
- 2024
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
Real-world intelligent systems often face the challenge of managing multiple data streams that experience concept drift, where the underlying data distribution changes over time. Hence, there's a rising interest in developing efficient learning techniques for analyzing multiple data streams with concept drift in non-stationary environments. The thesis offers an in-depth exploration of concept drift problem among multiple data streams, with a particularly focus on multistream classification. It provides a comprehensive definition and analysis of this problem, underscoring the significant challenges that dynamic and evolving data streams pose to intelligent decision-making systems. Therefore, this research proposes several innovative adaptive learning algorithms to enhance model performance for multiple non-stationary data streams. Key contributions include the Online Boosting Adaptive Learning (OBAL) method for dynamic correlation learning, the Fuzzy Shared Representation Learning (FSRL) approach for robust joint representation, the Learn-to-Adapt (L2A) framework for efficient high-dimensional data adaptation, and the Calibrated Source-Free Adaptation (CSFA) method for generalized class incremental learning. Additionally, this thesis further investigates the concept drift problem in a real-world application, focusing on video content restoration under concept drift conditions. It introduces a Transformer-based adaptive method to dynamically adapt the system's restoration capabilities in changing environments. To conclude, this research significantly advances adaptive learning under concept drift in multistream classification and real-world applications, improving system performance and adaptability in dynamic, non-stationary environments.
Please use this identifier to cite or link to this item: