The course is dedicated to deep learning techniques in time series modeling. Regardless of the task (forecasting, classification, etc.) deep learning models offer significant advantages when applied to time series problems. As there are no well established curriculum on deep learning for time series domain, the course is structured around a set of seminal research papers on the topic, several datasets and known use cases. Overall, it is covers main architectures and approaches, and designed to be practical.
The course does not require previous knowledge on the time series modeling, but requires some knowledge of deep learning concepts and toolkit.
Course prerequisites:
Robust knowledge of Python stack for data science and deep learning, incl. NumPy, Pandas, Jupyter, Matplotlib and PyTorch, general understanding of generic convolutional (CNN) and recurrent (RNN) neural net architectures, understanding of attention mechanism and transformer-like architectures will help, but is not required.