Project Details
Layman's description
Many applications in data science and machine learning are posed as optimization problems, one example is the supervised learning process of neural networks, where the goal is to minimize the error between the predicted and the expected output of the network. Several of these problems are sparse or have an inherent low-rank structure that can be exploited to design efficient algorithms.
Some problems such as semidefinite programming (SDP) leads to optimization with respect to dense matrices, even when the data matrices are sparse, which motivates the need to develop more complex methods to be possible to exploit sparsity in these cases. An example is the dimensionality reduction via semidefinite embedding, which can lead to an SDP problem with data matrices that have a hierarchical off diagonal low-rank structure.
The goal of this project is to construct new optimization methods that make use of rank structure, sparsity and hierarchical approximation techniques to improve performance and scalability. The project will explore different ways of utilize these techniques and analyse the trade-off between the efficiency and computational cost.
Some problems such as semidefinite programming (SDP) leads to optimization with respect to dense matrices, even when the data matrices are sparse, which motivates the need to develop more complex methods to be possible to exploit sparsity in these cases. An example is the dimensionality reduction via semidefinite embedding, which can lead to an SDP problem with data matrices that have a hierarchical off diagonal low-rank structure.
The goal of this project is to construct new optimization methods that make use of rank structure, sparsity and hierarchical approximation techniques to improve performance and scalability. The project will explore different ways of utilize these techniques and analyse the trade-off between the efficiency and computational cost.
Status | Active |
---|---|
Effective start/end date | 01/12/2022 → 30/11/2025 |
Fingerprint
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.