The Intelligent Robot Learning (IRL) Lab, affiliated with the University of Alberta and the Alberta Machine Intelligence Institute (Amii), has an immediate opening for a postdoctoral fellow to work with Matthew E. Taylor and multiple graduate students.
The goal of this multi-year project is to improve the way humans and reinforcement learning agents interact. In many cases, a human teacher may be able to provide additional guidance to the student agent. The goal of this feedback is typically to improve the agent’s learning (relative to learning without human guidance) while not harming the agent’s eventual performance (ensuring sub-optimal guidance does not hinder asymptotic performance). In particular, the lab considers questions such as:
– When and where is one type of human guidance more useful than another (e.g., a demonstration vs. providing positive and negative feedback)?
– What types of guidance would a human prefer to provide, and how does this interact with their perceived cognitive load?
– When should a computational agent ask for advice?
– Which of these methods, if any, are appropriate when the student is a human and the teacher is an agent?
The ideal candidate should have a PhD with demonstrated research experience in reinforcement learning, human-in-the-loop machine learning, explainability in machine learning, or human-robot interaction, with a strong record of publications. The candidate’s abilities in deep learning, python, human subject studies, and communication will also be considered.
The initial appointment will be for one year, is renewable subject to availability of funding and satisfactory performance, and offers a competitive salary.
The Computing Science department at the University of Alberta is an excellent and supportive environment of over 60 professors, 300 graduate students, and 1,300 undergraduate students. Matt’s research is situated inside of the larger Artificial Intelligence group with over 25 faculty members. Many are exploring issues in reinforcement learning, and others are interested in reinforcement learning interacting with humans. In addition, Matt is an Amii Fellow; Amii enriches the training environment by supporting dozens of students working in related areas, bringing high-profile visitors to the department, and supporting regular seminars for researchers to present their work.
UAlberta and IRL lab strives to provide equitable access to opportunities for post-doctoral fellows to disciplines of study and scholarly opportunities, while understanding this may require applying equitable effort; to study and disseminate knowledge about equity, diversity, and inclusivity; and to provide an environment attentive to, and that addresses, barriers to inclusion, access, and success (especially of historically excluded groups). Applicants from traditionally underrepresented groups are particularly encouraged to apply to this position. This position will provide the same family leave as is guaranteed to PhD students at the University.
To find out more or to apply, please send an email to Matt at mtaylor3+PDF2020@ualberta.ca with the subject “PDF2020”. If applying, please include a current CV and contact information for at least two references (cover letter optional).
View the full job posting here.