DTU - Postdoc in Geometric Machine Learning

Do you want to conduct free-form basic research that will allow humans to learn from the advances in machine learning?



DTU Compute invites applicants for a two-year Postdoc position in geometric machine learning.



If you have solid knowledge of Bayesian inference, geometry or representation learning and you are looking for an opportunity to put your research skills into practice you have it right here.



The Section for Cognitive Systems is looking to recruit a talented and highly motivated postdoctoral fellow for the ”Learning from Learned Representations” project, which is part of a bigger ambition to construct operational interpretable representations from unsupervised data analysis.



The project revolves around the goal of learning operational representations, i.e. representations that are naturally equipped with a set of well-defined operations that may be performed. For example, we may seek a representation that supports operators akin to addition and subtraction, or we may seek a representation that naturally supports integration (in order to assign probabilities to events). In practice, the project focuses on any aspect of the pipeline, ranging from Bayesian deep learning, deep generative models, random Riemannian geometry, numerical computations on manifolds, or other related topics. As such, the position is openly defined. 



For more details see: Operational representation learning. 



The Section for Cognitive Systems is an internationally renowned group for machine learning research. The group aims for the highest quality research. You're encouraged to collaborate both within the group and with other international groups. We emphasize a healthy work/life balance based on the premise that you do the best work when you are happy.



Responsibilities and tasks

Basically, your job will be to conduct free-form basic research at the highest international level. More specifically you will be building tools and theories for engage with identifiable representations. Doing so, requires building representation learning models that endow the representations with well-calibrated uncertainties, building computational tools for working with Riemannian representations, building theories on the interpretation of stochastic geometries, and many more such challenges. Your specific tasks depend on your background and interests. The position is flexible and will be adapted to the chosen candidate.



Most learned representations are treated as being Euclidean even if it is trivial to construct counter-examples showing that the Euclidean assumption lead to arbitrariness. You will join a team of people dedicated to avoiding this arbitrariness. You will work with nonlinear generative models and use geometric techniques to develop well-defined operations that can be meaningfully applied in the representation space of the model. The end-goal is to both improve the modelling capacity of generative models, but also to improve their general interpretability.



Depending on your interests and qualifications, the project can either be theoretical, applied, or a combination thereof. We generally believe that theory and applications must go hand in hand to ensure that the theory is meaningful and beneficial to scientific discovery.



More details are available at Open Positions.



Qualifications

You should have a PhD degree or equivalent in machine learning or a related field. You are expected to have both programming experience and be comfortable manipulating mathematical and probabilistic models. You will be working with a diverse group of people, so “people skills” are considered important.



Preference will be given to candidates with a publication record at top machine learning venues such as NeurIPS, ICML, ICLR, AISTATS, UAI, and similar.



We offer

DTU is a leading technical university globally recognized for the excellence of its research, education, innovation and scientific advice. We offer a rewarding and challenging job in an international environment. We strive for academic excellence in an environment characterized by collegial respect and academic freedom tempered by responsibility.



Salary and terms of employment

The appointment will be based on the collective agreement with the Danish Confederation of Professional Associations. The allowance will be agreed upon with the relevant union. The position is open from the end of 2022, but with a flexible starting date.



You can read more about career paths at DTU here.



Further information

Further information may be obtained from professor Søren Hauberg (sohau@dtu.dk). 



You can read more about the section for Cognitive Systems at www.compute.dtu.dk/english/research/Research-sections/Cogsys.



Application procedure

Please submit your online application no later than 15 October 2022 (Danish time). Applications must be submitted as one PDF file containing all materials to be given consideration. To apply, please open the link "Apply online", fill out the online application form, and attach all your materials in English in one PDF file. The file must include:

• Application (cover letter)
• CV
• Academic Diplomas (MSc/PhD – in English)
• List of publications

Applications and enclosures received after the deadline will not be considered.



All interested candidates irrespective of age, gender, disability, race, religion or ethnic background are encouraged to apply.

To apply please follow this link: Postdoc in Geometric Machine Learning (dtu.dk)

DTU Compute

DTU Compute is a unique and internationally recognized academic department with 385 employees and 11 research sections spanning the science disciplines mathematics, statistics, computer science, and engineering. We conduct research, teaching and innovation of high international standard – producing new knowledge and technology-based solutions to societal challenges. We have a long-term involvement in applied and interdisciplinary research, big data and data science, artificial intelligence (AI), internet of things (IoT), smart and secure societies, smart manufacturing, and life science. At DTU Compute we believe in a diverse workplace with a flexible work-life balance.



Technology for people

DTU develops technology for people. With our international elite research and study programmes, we are helping to create a better world and to solve the global challenges formulated in the UN’s 17 Sustainable Development Goals. Hans Christian Ørsted founded DTU in 1829 with a clear vision to develop and create value using science and engineering to benefit society. That vision lives on today. DTU has 13,400 students and 5,800 employees. We work in an international atmosphere and have an inclusive, evolving, and informal working environment. DTU has campuses in all parts of Denmark and in Greenland, and we collaborate with the best universities around the world.