KAKENHI(S) Discrete Structure Manipulation System Project (Hokkaido Univ.)

Sensor data analytics is one of the major application fields of data
mining and machine learning. Typically taking real-valued time-series
data from physical sensors as the input, its problem setting includes
a variety of tasks depending on application domains, not limited to
the traditional regression and classification. This talk will first
introduce technical challenges in industrial sensor data analytics.
Then it will cover recent developments in machine learning algorithms
in sensor data analytics. Major topics include change detection using
directional statistics and multi-task extension of graph-based
anomaly detection.