Tutorial-1
Machine Learning with Time Series Data
Ulsan National Institute of Science and Technology (UNIST), Korea
Abstract:
Time series models are essential to predict the future events in various applications such as finance, military and weather forecasting. This tutorial will first introduce traditional linear/non-linear dynamic models to analyze time-series data. This tutorial explain two recent advances in time series analysis: (1) Gaussian Processes based kernel decomposition method for time series data and (2) Deep Learning (Convolutional Neural Network) based method and its application on EEG (electroencephalogram) analysis.
Schedule:
14:00-15:00 Unit1. Basics of Dynamic Models
Contents: Kalman Filter, Extended KF, Unscented Kalman Filter ad Particle Filter
15:00-15:15 Break
15:15-16:15 Unit2. Gaussian Processes based methods
Contents: Gaussian Processes, Automatic Statistician
16:15-16:30 Break
16:30-17:30 Unit3. Deep Learning based methods
Contents: Recurrent Convolutional Layers, EEG data analysis
Bio:
Education
May 2012: PhD, Computer Science, University of Illinois at Urbana-Champaign
August 2004: BS, Computer Engineering, Seoul National University
Positions
July 2013 –Present: Assistant Professor, School of Electrical and Computer Engineering, Ulsan National Institute of Science and Technology
January 2013 – July 2013: Postdoc, Lawrence Berkeley National Laboratory
May 2012– January 2013: Postdoc, University of Illinois at Urbana-Champaign
Selected Research Publications
The Relational Automatic Statistician, International Conference on Machine Learning, 2016
Spatio-Temporal Pyramid Matching, Computer Vision and Image Understanding, 2013
Lifted Relational Kalman Filtering, International Joint Conference on Artificial Intelligence, 2011
Research Interests
Statistical Inference, Probabilistic Graphical Model, Dynamic Bayesian Models, Robot Task Planning