Decoupling preconditioners for non-smooth optimisation and inverse problems
Description of the granted funding
The solution of inverse and learning problems is crucial in many modern technologies, such as medical and astronomical imaging devices, and personal assistants. As the physical technologies improve, and become more widespread, we are faced with increased amounts of data, as well as the need to process them into higher-quality results. We would also like to use as little energy as possible, and get results in real time. For this we need very good optimisation algorithms: instructions for a computer to find the point where a mathematical function reaches its smallest value, akin to the bottom of a valley among hills. This project will develop such efficient optimisation algorithms by carefully exploiting all the structural information available, yet cheaply. Instead of climbing the steepest slope of a hill, which maybe be the shortest but a very energy-intensive path, it can be overall cheaper to go many times around the hill - if the overall landscape looks promising.
Show moreStarting year
2018
End year
2024
Granted funding
Related funding decisions
345389
Research costs of Academy Research Fellows(2021)
159 852 €
320022
Research costs of Academy Research Fellows(2018)
209 881 €
Funder
Research Council of Finland
Funding instrument
Academy research fellows
Other information
Funding decision number
314701
Fields of science
Mathematics
Research fields
Sovellettu matematiikka
Identified topics
computer science, information science, algorithms