Lund University, Faculty of Engineering, LTH, Department of Computer Science

Lund University was founded in 1666 and is repeatedly ranked among the world’s top 100 universities. The University has 40 000 students and 7 600 staff based in Lund, Helsingborg and Malmö. We are united in our efforts to understand, explain and improve our world and the human condition.

LTH forms the Faculty of Engineering at Lund University, with approximately 9 000 students. The research carried out at LTH is of a high international standard and we are continuously developing our teaching methods and adapting our courses to current needs.

Subject description
Black box optimization concerns the optimization of functions that can only be evaluated through numerical simulation and in which partial derivates are either not known or not defined. One way to optimize these functions consists of using consecutive function evaluations to build and refine a surrogate model, and then use this model to drive the optimization. Albeit very effective, this model-based approach is typically computationally expensive, becoming impractical when the number of input variables is greater than a few dozens and when multiple objectives are considered jointly.

This project, financed by WASP (Wallenberg AI, Autonomous System and Software Programme, http://wasp-sweden.org), aims at introducing innovative algorithms and methodologies to overcome the limitations of multi-objective black-box optimization. The research topic falls at the crossroads of the broader fields of black-box optimization and statistical machine learning. This project will develop statistical methods for modelling surrogate models. These models can then be queried using Bayesian optimization or similar techniques to identify values of process settings that optimize the quality of the multiple objectives, or that maximize the information gained from experiments. Using a Bayesian statistical approach augmented with user prior knowledge will enable combining information from different sources, e.g. experiments of different kinds, while accounting for the uncertainty involved in a rigorous and coherent manner. The new algorithms and methodologies will be tested on a variety of synthetic and real-world applications such as automated machine learning (AutoML), automated configuration of compilers, hardware design and computer vision.

This project is part of a collaboration with Stanford University. The student will be encouraged to apply to the WASP exchange program with Stanford to work closely with collaborators.

Topics of interest:

  • Black-box optimization.
  • Derivative-free optimization (DFO).
  • Bayesian optimization.
  • Algorithm configuration and selection.
  • Active learning.
  • Automated machine learning (AutoML).
  • Neural architecture search (NAS).
  • Hyperparameter optimization.
  • Learning to learn.
  • Meta learning and transfer learning.
  • Reinforcement learning (RL).
  • Optimization of neural networks.
  • Evolutionary algorithms (EA).
  • Discrete optimization and NP-hard problem solving.
  • Data-driven analysis of algorithms, hyperparameter importance, etc.

Some applications of interest:
Image classification. Natural Language Processing (NLP). Simultaneous localization and mapping (SLAM). Design space exploration (DSE). Optimizing compilers. Hardware design: CPU, GPU, FPGA, CGRA, ASIC.

Work duties
The Department of Computer Science is looking for one or two doctoral students to this project. The main duties of doctoral students are to devote themselves to their research studies which includes participating in research projects and third cycle courses. The work duties can also include teaching and other departmental duties (no more than 20%).

The research involves the following subtasks:

  • Review of available scientific literature of the research topic.
  • Design theoretical frameworks, methodologies and their implementations.
  • Run experiments on synthetic and real-world applications using a principled empirical evaluation.
  • Be present and participate in the research and education environment of the research group by actively taking on duties.
  • Communication of the obtained results in the forms of publication of scientific articles and talks in top tier venues.

Admission requirements

A person meets the general admission requirements for third-cycle courses and study programmes if he or she:

  • has been awarded a second-cycle qualification, or
  • has satisfied the requirements for courses comprising at least 240 credits of which at least 60 credits were awarded in the second cycle, or
  • has acquired substantially equivalent knowledge in some other way in Sweden or abroad

A person meets the specific admission requirements for third cycle studies in Machine Learning (Computer Science or Mathematics) if he or she has:

  • at least 150 credits in mathematics, engineering and science including at least 60 second-cycle credits in computer science or mathematics and a second-cycle degree project worth 30 credits of relevance to computer science or mathematics, or
  • a MSc in Engineering including at least 60 credits in computer science or mathematics, or another second-cycle qualification of relevance to machine learning, including at least 60 credits in computer science or mathematics.

Additional requirements:

  • Very good oral and written proficiency in English.

Assessment criteria
Selection for third-cycle studies is based on the student’s potential to profit from such studies. The assessment of potential is made primarily on the basis of academic results from the first and second cycle. Special attention is paid to the following:

1. Knowledge and skills relevant to the thesis project and the subject of study.
2. An assessment of ability to work independently and to formulate and tackle research problems.
3. Written and oral communication skills
4. Other experience relevant to the third-cycle studies, e.g. professional experience.

Other assessment criteria:

  • Bachelor/master thesis in statistical learning and/or computer vision, with focus on Black-box optimization and AutoML.
  • Good programming skills in Python or similar, a deep knowledge of compiler technology and hardware accelerators are considered a merit.

Consideration will also be given to good collaborative skills, drive and independence, and how the applicant, through his or her experience and skills, is deemed to have the abilities necessary for successfully completing the third cycle programme.

Terms of employment
Only those admitted to third cycle studies may be appointed to a doctoral studentship. Third cycle studies at LTH consist of full-time studies for 4 years. A doctoral studentship is a fixed-term employment of a maximum of 5 years (including 20% departmental duties). Doctoral studentships are regulated in the Higher Education Ordinance (1993:100), chapter 5, 1-7 §§.

Instructions on how to apply
Applications shall be written in English and include a cover letter stating the reasons why you are interested in the position and in what way the research project corresponds to your interests and educational background. The application must also contain a CV, degree certificate or equivalent, and other documents you wish to be considered (grade transcripts, contact information for your references, letters of recommendation, etc.).

Type of employment Temporary position longer than 6 months
First day of employment As soon as possible
Salary Monthly salary
Number of positions 2
Working hours 100
City Lund
County Skåne län
Country Sweden
Reference number PA2019/2533
Contact
  • Jacek Malec, jacek.malec@cs.lth.se
  • Luigi Nardi (questions about the project), luigi.nardi@cs.lth.se
Published 05.Jul.2019
Last application date 29.Jul.2019 11:59 PM CET

Return to job vacancies