Online applications are invited for the Visiting Student Research Program at KAUST. Check the details and Apply Now!
About KAUST
KAUST (King Abdullah University of Science and Technology) aspires to be a destination for scientific and technological education and research. By inspiring discoveries to address global challenges, we strive to serve as a beacon of knowledge that bridges people and cultures for the betterment of humanity.
About Visiting Student Research Program
The Visiting Student Research Program (VSRP) is an exciting internship opportunity aimed at offering research experience for highly qualified and motivated international students in science, technology, engineering, and mathematics. Collaborate with their faculty in world-class facilities. Contribute to new knowledge that’s not only published in journals but also helps foster solutions that impact society.
Roles and Responsibilities
- Project Title: Parameter – Free Optimization, Universal Prediction, and Kolmogorov Complexity.
- Recommended Student Academic and Research Background:
- Knowledge of convex optimization
- Knowledge of Kolmogorov complexity
Eligibility
- Program – Computer Science
- Division – Computer, Electrical and Mathematical Sciences and Engineering
- Faculty Lab Link – Visit here
- Field of Study –Computer Science, Mathematics, or a related discipline
Project Description
- Online parameter-free optimization algorithms are optimization algorithms that do not require tuning of parameters, yet they achieve optimal performance.
- The basic idea behind these algorithms is the link between universal strategies for prediction with log loss and online convex optimization, first unveiled by Orabona and Pal (2016).
- By now, this line of work has produced several parameter-free optimization algorithms. However, universal prediction with log loss is also connected to the concept of Kolmogorov complexity for binary strings, as shown by Cover (1974).
- In this project, they aim to study all these links, going from parameter-free optimization to Kolmogorov complexity, passing through universal prediction with log loss.
- They want to construct explicit reductions to transform each of these problems into one of the other ones.
- The final aim is to directly link convex optimization to an appropriate notion of computational complexity for functions. Moreover, they want to extend some of these concepts from binary strings to strings of bounded real numbers.
- Given the theoretical nature of this project, the ideal candidate must have an excellent mathematical background.
How to Apply?
Interested Candidates may apply here.
Salary
INR 7 LPA
