Column reduced digital nets

Area of Science:

• Numerical Analysis
• Computational Mathematics
• Quasi-Monte Carlo Methods

Background:

• Digital nets efficiently generate integration nodes for quasi-Monte Carlo (QMC) rules.
• Accelerating matrix-vector products with QMC nodes is crucial for applications like uncertainty quantification.
• Prior work utilized reduced lattices and row reduced digital nets for speed-up.

Purpose of the Study:

• To develop a novel matrix-vector multiplication algorithm for QMC.
• To exploit the repetitive structure of column reduced digital nets.
• To improve upon existing methods for fast QMC matrix-vector products.

Main Methods:

• Proposed a new multiplication algorithm leveraging column reduced digital nets.
• Developed an upper bound for the quality parameter of column reduced digital nets.
• Conducted numerical tests to validate the algorithm's efficiency.

Main Results:

• The new algorithm exploits the structure of column reduced digital nets for faster matrix-vector products.
• This method facilitates error analysis compared to previous approaches.
• Numerical tests demonstrate the efficiency of the proposed algorithm.

Conclusions:

• The developed algorithm offers an efficient alternative for matrix-vector products in QMC.
• Exploiting column reduced digital nets provides advantages in speed and error analysis.
• The findings contribute to the advancement of computational methods in QMC applications.