Time-series regression with deep neural networks

About this talk

Recent progress in real-valued time-series crush the state-of-the-art of autoregressive models in many domains like neuroscience, stock markets and energy. I collected some of the most interesting models, like seq2seq, neural processes and GANs to make the research more accessible. During the talk we will walk trough to the challenges of real-valued time-series regression, the models that are currently used in state-of-the-art research and their novelties. Finally we will code a simple simple use case about price forecasting for cryptocurrencies.