Many academic departments have heavy teaching.
Teaching is intense and students depend on us.
Teaching have shorter time scales than research. Crowding out.
Teaching culture can quickly become dominant.
How to develop a research culture and maintain teaching efforts?
OED: a way of life or social environment characterized by or associated with the specified quality or thing.
Naturally occurring research discussions at the coffee machine.
Synergistic relationship between teaching and research?
Is it possible to develop a culture, or must it happen organically?
Connect junior researcher to senior mentor for a grant proposal.
Grant-pitching seminars for feedback and a senior connection.
Feedback throughout the process, particularly on “the big picture”.
Plus/Minus: Mentor may end up as co-applicant.
Co-applicant in established research group, also in other subjects.
“Consulting” towards other departments to get contacts.
Suitable offices and equipment for thinking and creating.
Library and database access and all of that.
Key physical books in smaller department library.
Brainstorming rooms with big display, whiteboards, comfy chairs.
Whiteboards in the coffee room for spontaneous discussions.
New books (research monographs) on display in coffee room.
Bulletin board with
Learning about new research.
Learning about the subject, both broad and deep.
Research trends - important for less active researchers.
Meet other researchers.
Fund the seminars
Attract speakers from abroad.
Get speakers to stay for some days.
Schedule one-on-one meetings with speaker.
Also with PhD students and less research active staff.
Make seminars visible on a public web page, incl past seminars.
Link to past seminars when inviting speakers. Stars attract stars.
When paper is available, distribute it to staff for better discussion.
PhD course credit for acting as unofficial mini-discussant.
Distribute the invitation and hosting across all researchers.
Make it the department’s seminar series.
We recently got the paper ‘Real-Time Robotic Search using Hierarchical Spatial Point Processes’ accepted for publication in the machine learning conference ‘Uncertainty in Artificial Intelligence’ (UAI2019).
The paper is joint work with my PhD students Olov Andersson (Robotics) and Per Sidén (Statistics), my previous post doc Johan Dahlin (Automatic Control) and Prof Patrick Doherty (Robotics). It is funded by Patrick’s and my SSF project in Smart Systems.
Abstract: Aerial robots hold great potential for aiding Search and Rescue (SAR) efforts over large areas. Traditional approaches typically searches an area exhaustively, thereby ignoring that the density of victims varies based on predictable factors, such as the terrain, population density and the type of disaster. We present a probabilistic model to automate SAR planning, with explicit minimization of the expected time to discovery. The proposed model is a hierarchical spatial point process with three interacting spatial fields for i) the point patterns of persons in the area, ii) the probability of detecting persons and iii) the probability of injury. This structure allows inclusion of informative priors from e.g. geographic or cell phone traffic data, while falling back to latent Gaussian processes when priors are missing or inaccurate. To solve this problem in real-time, we propose a combination of fast approximate inference using Integrated Nested Laplace Approximation (INLA), and a novel Monte Carlo tree search tailored to the problem. Experiments using data simulated from real world GIS maps show that the framework outperforms traditional search strategies, and finds up to ten times more injured in the crucial first hours.
Computer Science has a completely different publication culture than Statistics, where the majority of publications are 6-10 pages papers in conference proceedings. The conferences are ranked and some of the conferences have higher status than the best journals. So, unlike Statistics, conference proceedings really matters. The top general conferences in machine learning are NeurIPS and ICML. Those conferences are ranked A+. UAI is ranked A, so it is a very good conference specialized in probabilistic modeling (mostly graphical models) for AI. The accept rate for UAI this year was 26% (compared to 20% for NeurIPS).
It is common in Robotics to complement the experiments in the paper with a video. Here is our video for the paper: https://www.youtube.com/watch?v=wyD0O5hF5tE.