Large-Scale Optimization

The WASP Core Technology Cluster: Large–Scale Optimization, is a platform for WASP students with an interest in optimization. Its main purpose is to enable students that research optimization theory and its applications to exchange ideas, network, and collaborate. We arrange activities continuously to achieve this. The spectrum of topics we discuss ranges from theoretical aspects of optimization frameworks to the development of novel optimization schemes tailored for various applications in e.g. machine learning, computer vision, and control. To name a few topics, we research:

  • development of novel algorithms for scalable machine learning,
  • distributed optimization,
  • performance estimation of algorithms,
  • applications to representation learning,
  • etc.

Some activities we arrange include:

  • paper sessions,
  • presentations on research projects (external and internal)
  • meet-ups.

 

Cluster Leader

Jacob Lindbäck

PhD (WASP AI), School of Electrical Engineering and Computer Science (EECS), KTH

Send Email

Cluster Members

Affiliation: WASP Ph.D. student at KTH, EECS
Main Supervisor: Mikael Johansson
Keywords: scalable optimization for machine learning, distributed optimization.
Research Abstract: My research is mainly concerned with developing and establishing theoretical guarantees for novel algorithms tailored for various learning tasks on distributed architectures. For the time being, I am researching ADMM based approaches to solve data integration tasks in biostatistics.

Affiliation: WASP Ph.D. student at Chalmers University of Technology, Department of Electrical Engineering
Main Supervisor: Prof. Christopher Zach
Keywords: geometric computer vision, representation learning, 3D vision
Research Abstract: Learning and Leveraging Rich Priors for Factorization Problems — we aim to combine traditional parametric formulations induced by domain expertise with non-parametric models learned from examples; main application of interest are factorization-based problems, in particular non-rigid structure-from-motion.

Affiliation: WASP Ph.D. student at Lund University
Main Supervisor: Pontus Giselsson
Keywords: convex optimization, performance estimation problems.
Research Abstract: Current research interest is within the analysis and design with respect to worst-case performance of iterative algorithms for optimization and inclusion problems.

–

–

–

–

–

–

–

–

–

–

–

–

Cluster activities thus far

  • Paper Session (December 14th), coordinated by Yaroslava Lochman
    Paper: Learning 2D-3D Correspondences to Solve the Blind Perspective-n-Point Problem, L. Liu, D. Campbell et al.
  • Research presentation (November 8th) by Jacob Lindbäck
    Jacob Lindbäck presented his ongoing research projects.
    Title: Splitting methods for OT and Manifold Optimization
  • Research presentation (September 17th) by Yaroslava Lochman
    Yaroslava Lochman gave a presentation about her current research project.
    Title: Learning Rich Priors for Factorization Problems
  • Paper Session (June 11th), coordinated by Manu Upadhyaya
    Paper: Performance of first-order methods for smooth convex minimization, Y. Drori  & M. Teboulle
  • Paper Session  (April 27th), coordinated by Jacob Lindbäck
    Paper: A Fast Iterative Shrinkage-Thresholding Algorithm for Linear Inverse Problems, A. Beck & M. Teboulle