Machine learning for behavioural modelling
Lead Research Organisation:
University of Oxford
Department Name: Engineering Science
Abstract
This project is a an ICASE studentship supported by Dyson Technology Ltd supervised by Dr Michael Osborne.
We are studying the extraction of behavioural patterns from corpora of past interaction histories with unknown agents via machine learning methods in general and Bayesian reasoning in particular. Based on such learned patterns we will then explore new means of optimising future interactions with these agents.
A second point of interest, is research regarding mechanisms for correcting derived models based on specific evidence which suggests that the behaviours in the set do not approximate the real agent types well enough.
To ensure the robustness of the developed algorithms, we will conduct eventual experiments on data sets provided by Dyson, build from real-world customer data. As a consequence, our models will need to be able to adapt to uncontrollable and changing environments such as for instance domestic settings.
By pursuing this research, we hope to contribute to the development of machines capable of truly autonomous behaviour, able to assess their environment, make decisions, and learn from the results which is a major challenge in practical robotics.
This project falls within the EPSRC information and communication technologies research area.
We are studying the extraction of behavioural patterns from corpora of past interaction histories with unknown agents via machine learning methods in general and Bayesian reasoning in particular. Based on such learned patterns we will then explore new means of optimising future interactions with these agents.
A second point of interest, is research regarding mechanisms for correcting derived models based on specific evidence which suggests that the behaviours in the set do not approximate the real agent types well enough.
To ensure the robustness of the developed algorithms, we will conduct eventual experiments on data sets provided by Dyson, build from real-world customer data. As a consequence, our models will need to be able to adapt to uncontrollable and changing environments such as for instance domestic settings.
By pursuing this research, we hope to contribute to the development of machines capable of truly autonomous behaviour, able to assess their environment, make decisions, and learn from the results which is a major challenge in practical robotics.
This project falls within the EPSRC information and communication technologies research area.
Organisations
People |
ORCID iD |
Michael Osborne (Primary Supervisor) | |
Sebastian Schulze (Student) |
Publications
Nguyen V
(2020)
Bayesian Optimization for Iterative Learning
Zintgraf L
(2020)
VariBAD: A Very Good Method for Bayes-Adaptive Deep RL via Meta-Learning
Studentship Projects
Project Reference | Relationship | Related To | Start | End | Student Name |
---|---|---|---|---|---|
EP/P510609/1 | 30/09/2016 | 29/09/2021 | |||
1802029 | Studentship | EP/P510609/1 | 30/09/2016 | 29/06/2021 | Sebastian Schulze |