No openings at the moment.¶
There was a recent opening for two PhD positions. The application window has closed and we cannot accept further applications. Keeping the text below for a short time for reference.
I have an opening for two PhD candidates in Machine Learning for Natural Language Processing.
Positions are fully funded for four years.
The call is at this link, alongside instructions to apply. Please apply via that link by Aug 21 2022.
Emailing me: Don’t email me to apply: you must use the website or else I can’t consider you. Do email me with any specific questions — please include [PhD 9665] in the subject line so my filters can catch your email.
Both positions are on deep generative models of language.
One of the positions, connected to the Hybrid Intelligence project, is oriented toward controllability, constraints, and geometry in deep generative models of language via continuous representations, taking cue from recent developments in continuous energy-based models and diffusion models. We will be particularly looking into continuous models of language, to overcome the tension between the discreteness of language and the continuity of deep learning, taking cue from recent developments in continuous energy-based models and diffusion models.
- Representing, quantifying, and using uncertainty in generative models of language, conversations, or other linguistic structures;
- Explainability methods for conditional language generation, in particular with long documents and contexts;
- Adaptable, multi-modality aware generation, responsive to context, constraints and preferences.
Please pick one or two of these directions and express and justify your interest in it in application letter. A strong application should also demonstrate awareness of some of the recent work in these directions, and should briefly describe research ideas in relationship to existing work.
We will rely on a selection of machine learning techniques. My research gravitates around structured prediction, latent variable models, sparsity, convex and constrained optimization, and geometry (machine learning on manifolds.) Importantly, You do not have to already be an expert in these ML topics. But if you have ML basics and you are enthusiastic to study one or more of these (or related) topics, together with me, in order to apply them in NLP research, you are a perfect fit!
The flavor of my research is to integrate structured and discrete representations into the “soft” computation of neural networks. See this captured in my recent papers.
- The master’s degree requirement is unfortunately firm, and out of my control.
- We are looking for a starting date of Sept-Dec 2022. Please specify your availability in the application.
- In your letter, focus on what you want to do in your PhD, not as much on what you have done in the past (especially if your past work is not on the very specific topics in this call.)
- If you want to do your master’s at the UvA and work with me, emailing me does not help; you must apply to the program. At UvA, master’s students usually are not attached to research labs, but me or my LTL colleagues can advise your thesis project in the second year.
- I don’t at the moment have the ability or the capacity to recruit interns or self-funded PhD students.
It’s a good idea to be skeptical about working with an advisor who is early in their career. Talking to their other students is always good: I am currently advising Evgeniia Tokarchuk at UvA, and have co-advised Tsvetomila Mihaylova and Gonçalo M. Correia. You can also write to me with any questions.
The PhD is a job, and, at UvA, PhD students are employees, with the workers’ rights that stem from this. I believe your PhD is not your entire life and identity. I encourage and try to actively demonstrate work-life balance, awareness and support of mental health struggles, inclusivity, openness and safety. The PhD is the time to learn how to learn, ask, answer, describe, and teach. I will help and mentor you through it, according to my ability and your requested level.