ESR8. Self-adaptation and auto-configuration of probabilistic graphical models for SH and AAL applications
Project Title: Self-adaptation and auto-configuration of probabilistic graphical models for SH and AAL applications.
Description: Project will mainly focus on dynamic dialog systems, capable of practically be deployed on real world applications, through avatars to increase the quality of smart home avatars designed for older persons with cognitive impairments.In our proposal we will be addressing the issue of dialogue creation tool that is needed to create samples to train reinforcement learning and also focus on more sample-efficient and adaptive reinforcement learning methods. Since people with cognitive impairments tend to change their behavior in course of time and their preferences, it is also crucial to develop a reinforcement learning algorithm which will take into account temporal changes on problem space for same user. (Action, state, reward) Automated creation of dialogues and more efficient and intelligent systems would make it possible to use dynamic dialogue systems to be used in smart home applications for people with dementia.
ESR8: Erinc Merdivan
Affiliation: Austrian Institute of Technology Vienna