Key research themes
1. How do active learning algorithms improve sample efficiency and robustness in practical and realistic data scenarios?
This theme focuses on the development of active learning frameworks and algorithms that address challenges in real-world data, such as class imbalance, out-of-distribution (OOD) samples, data redundancy, and labeling cost. It is crucial for deploying active learning in practical applications where data is vast and noisy, and labeling resources are constrained.
2. What are the design principles and performance trade-offs of active filter topologies for power quality and harmonic mitigation in electrical systems?
This theme investigates the architectures, control strategies, and comparative advantages of various active filter designs used to improve power quality by mitigating harmonics, voltage sags, swells, and reactive power in electrical distribution systems. Understanding these designs is essential for developing efficient power conditioners in industrial and grid applications.
3. How can adaptive filter algorithms and their combinations enhance tracking and convergence performance for time-varying systems efficiently?
This theme delves into the theory and application of adaptive filtering methods, including combinations of LMS and RLS filters, aimed at accurate and fast-converging estimation in non-stationary and time-varying environments. It focuses on achieving near-optimal tracking performance with minimal computational complexity, which is critical for real-time signal processing and control in dynamic systems.