Advice for Prospective PhD Students
As faculty at The University of Edinburgh, I have begun to receive many emails from prospective students asking for advice on applying to do a PhD with me. I'm often unable to answer all of these, but here are some answers to many commonly asked questions.
Application Process
For PhDs at Edinburgh in 2024, you'll need to apply via "EUCLID". You can apply to the Institute for Perception, Action, and Behaviour (IPAB) here
When to Apply (and when to Start)
Most students start their PhD in September. (Note: Technically, January/May starts are also possible...but this is tricker, as most internal selection timelines operate assuming a September start). Remember - regardless of start date, you'll need sufficient time to go through the full application process.
The official policy on this is, in theory: For UK students, apply at minimum a month before your desired start date. For International Students, 3 or 4 months beforehand (because of the extra bureaucracy around Visa Requirements, ATAS, CAS etc.,).
In practice however, you may want to consider applying much earlier. This is because often particular funding opportunities are tied to much earlier deadlines.
For example, to be considered for the Informatics Global PhD Studentship in the 2024 period, the two application deadlines are 24th Nov 2023 and 26 Jan 2024.
More info on application timings can be found here
Funding the PhD
There are a number scholarship types for funding a PhD. Eligibility will depend on whether you are a UK, EU, or International Student.
You will be considered for some of these funding sources automatically if your application arrives early enough. For example, the Informatics Global Scholarships (open to International students), or the EPSRC Doctoral Training Partnerships (open to Home/UK Students).
Other funding sources require separate application phases. For an overview of available scholarship sources, see this page.
Centres for Doctoral Training
An alternative funding route are the Centres for Doctoral Training. These are 4-year degrees. The first year consists of taught courses, research skills and project formulation; The remaining years constitute a standard PhD path.
Different centres have different themes. If you are working with me, the most relevant CDT will be the CDT on Dependable Robotics and Autonomous Systems.
Project Specific Funding
Some supervisors have individual funding for PhD topics in a specific interest area. For example, I have an open position on Hybrid AI for Cyber-Physical Systems.
Other projects can be found at Project related PhD Studentships.
Coming up with a Proposal
Your research proposal (max two pages) is the most important part of your application. It doesn't pin you down to what you will study during the PhD (it may change), but shows you have an idea for an interesting, feasible problem. The University of Edinburgh has a guide on How to write a good research proposal. So does FindAPhD.
For starting ideas, read the work of a potential supervisor. You can find a listing of all a supervisor's publications via Google Scholar. Here is my profile. I am interested in:
- Integrating ML systems for perception/control with temporal, spatial and temporal logics.
- Extracting rules and scenario specifications from requirements documents (e.g., Highway traffic rules, crash incident reports), for use in probabilistic configuration languages.
- Developing tools to analyze & calibrate how risk propagates across ML modules.
- Using statistical experiment design to improve efficiency and coverage of ML-based simulations.
Another idea: Read a Survey paper in an area you are interested in. Try to identify a research question in the literature. Here are some examples to start you off:
- A Survey on Mining Signal Temporal Logic Specifications
- Survey on Scenario-Based Safety Assessment of Automated Vehicles
- A Survey of Algorithms for Black-Box Safety Validation of Cyber-Physical Systems
- A Survey of Statistical Model Checking
- Formal methods to comply with rules of the road in autonomous driving: State of the art and grand challenges