Recurrent neural networks for structured data

A key challenge in machine learning is to explore and incorporate the complex nature of real-world data structures into the training models. The contributions of this thesis are novel RNN architectures for different types of structured data.