Capturing Regular Human Activity through a Learning Context Memory

Philipp H. Mohr, Nick Ryan, Jon Timmis

A learning context memory consisting of two main parts is presented. The first part performs lossy data compression, keeping the amount of stored data at a minimum by combining similar context attributes --- the compression rate for the presented GPS data is 150:1 on average. The resulting data is stored in an appropriate data structure highlighting the level of compression. Elements with a high level of compression are used in the second part to form the start and end points of episodes capturing common activity consisting of consecutive events. The context memory is used to investigate how little context data can be stored containing still enough information to capture regular human activity.

Subjects: 12.1 Reinforcement Learning; 11. Knowledge Representation

Submitted: May 31, 2006

This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.