Multi-tasking in data science: recent advances and open questions
|
Organizer(s): |
Name:
|
Affiliation:
|
Country:
|
Angelica Aviles-Rivero
|
University of Cambridge
|
England
|
Rihuan Ke
|
University of Cambridge
|
England
|
Carola-Bibiane Schönlieb
|
University of Cambridge
|
England
|
|
Abstract:
| | Multi-tasking (MT) denotes the approach of jointly carrying out a list of tasks, and is an important topic in various areas of data science. The MT strategy takes advantage of dependencies between tasks and by treating them jointly either (i) improves the solution quality of one or all tasks by exploiting this correlation in the MT model, or (ii) facilitates the solution of one task by linking it to another one. Methodological frameworks of the MT idea include multi-task and transfer learning, as well as variational regularisation and constrained optimisation approaches, with application areas in weak labelling, task-adapted image enhancement, feature reconstruction in computed tomography, fast imaging, and many more. In this special session we will present state-of-the-art approaches, theory and applications, as well as discuss mathematical and practical challenges ahead. |
|
|
List of approved abstract |
|
|
|
|
