US Patent Application 17657912. SYSTEMS AND METHODS FOR SPARSE MATRIX MULTIPLICATION simplified abstract

From WikiPatents
Jump to navigation Jump to search

SYSTEMS AND METHODS FOR SPARSE MATRIX MULTIPLICATION

Organization Name

Microsoft Technology Licensing, LLC==Inventor(s)==

[[Category:Venmugil Elango of Redmond WA (US)]]

[[Category:Bita Darvish Rouhani of Bellevue WA (US)]]

[[Category:Eric S Chung of Woodinville WA (US)]]

[[Category:Douglas Christopher Burger of Bellevue WA (US)]]

SYSTEMS AND METHODS FOR SPARSE MATRIX MULTIPLICATION - A simplified explanation of the abstract

This abstract first appeared for US patent application 17657912 titled 'SYSTEMS AND METHODS FOR SPARSE MATRIX MULTIPLICATION

Simplified Explanation

The patent application describes a method for multiplying sparse matrices efficiently.

  • The method involves receiving a first block of elements and dividing it into smaller sub-blocks with a certain level of sparsity.
  • A sparsity mask is applied to the first block to ensure that each sub-block has the same level of sparsity.
  • A second block of elements is received and divided into sub-blocks with a different level of sparsity.
  • A sparsity mask is applied to the second block to ensure that a certain percentage of sub-blocks have no elements and the remaining sub-blocks have all elements.
  • Finally, the first and second blocks are multiplied together using matrix multiplication.


Original Abstract Submitted

A method for sparse matrix multiplication comprises receiving a first block having M elements in a first dimension, and parsing the first block of M elements into a first set of B sub-blocks including MB elements in the first dimension. A first sparsity mask having S % sparsity is applied to the first block of elements, such that each of the first set of B sub-blocks has S % sparsity. A second block is received having M elements in a second dimension, and is parsed into a second set of B sub-blocks that include MB elements in the second dimension. A second sparsity mask having S′% sparsity is applied to the second block of elements, such that S′% of the second set of B sub-blocks have 100% sparsity and (100−S′)% of the second set of B sub-blocks have 0% sparsity. The first and second blocks are then matrix multiplied.