Summary: Real-world problems are often characterized by data arising from non-stationary environments, with data evolving over time. These problems can only be addressed by algorithms that are capable of learning from environments without a priori assumption of the data, and without the need to retrain the network as new data arrives. Such algorithms also empower a learning system with increased intelligence to retain their knowledge and adapt themselves for further learning tasks.

This talk will present a broad overview of the recent advancements in online learning algorithms, and their application to real-world problems involving multi-variate time series data. In addition, the talk shall also briefly walk through my research on complex-valued neural networks.