Time series modeling with deep learning

A major part of real-world phenomena and processes, such as speech, text, financial assets, DNA, etc., can be described as sequential data or time series. These data are often loaded with additive noise and may also follow sometimes chaotic behaviour, thus, the signal-to-noise ratio may be very low. There have been numerous approaches to model sequential data in the last decades. The rise of deep learning resulted in novel techniques that are likely to give better results than prior methods. A major part of real-world phenomena and processes, such as speech, text, financial assets, DNA, etc., can be described as sequential data or time series. These data are often loaded with additive noise and may also follow sometimes chaotic behaviour, thus, the signal-to-noise ratio may be very low. There have been numerous approaches to model sequential data in the last decades. The rise of deep learning resulted in novel techniques that are likely to give better results than prior methods.
In this talk, we will discuss the basic and state-of-the-art time series modeling techniques with deep learning and will also discuss the common pitfalls.

Gyires-Tóth Bálint
Researcher, Senior Lecturer, BME TMIT

Bálint Gyires-Tóth conducts research on fundamental and applied machine learning since 2007. With his leadership, the first Hungarian hidden Markov-model based Text-To-Speech (TTS) system was introduced in 2008. Since 2014 his primary research field is deep learning, focusing on sequential data modeling with deep learning and deep reinforcement learning. He was involved in various successful research and industrial projects. He is a certified NVidia Deep Learning Institute (DLI) Instructor and University Ambassador.