Skill discovery in reinforcement learning

A fundamental problem in Artificial Intelligence (AI) is how to endow artificial agents with the ability to autonomously form useful high-level behaviors (for example, grasping) from available behavioral units (for example, primitive sensory and motor actions available to a robot). This ability allows a developmental process during which an agent can learn to display behaviors of increasing complexity through continuously building on its existing set of skills to acquire new ones. For example, grasping may be followed by manipulating objects in different ways, which may be followed by using a key to unlock the door to an adjacent room, and so on, forming a continuously growing hierarchy of behaviors.

This project aims to develop algorithms that enable an artificial agent to go through such a developmental process. During this process, the agent will develop the capability of performing tasks of increasing difficulty without guidance tailored for the specific domain in which it operates. This type of open-ended developmental process is fundamentally different from how artificial agents learn to perform complex tasks today and has the potential to dramatically increase their capabilities.

Specifically, the project aims to develop algorithms for autonomously identifying and learning useful high-level behaviors as an agent interacts with its environment, driven by curiosity, using the well-developed framework of reinforcement learning.

Desirable qualities in candidates include intellectual curiosity, a strong background in math and extensive programming experience.

Applicants should hold, or expect to receive, a First Class or good Upper Second Class Honours degree, or the equivalent from an overseas university. A master’s level qualification would also be advantageous.

Informal enquiries should be directed to Dr Özgür Şimşek: [Email Address Removed]

Formal applications should be made via the University of Bath’s online application form for a PhD in Computer Science: https://samis.bath.ac.uk/urd/sits.urd/run/siw_ipp_lgn.login?process=siw_ipp_app&code1=RDUCM-FP01&code2=0013

Start date: 30 September 2019.

Note: Applications may close earlier than the advertised deadline if a suitable candidate is found; early application is therefore recommended.