Matrix factorization with two-stage data block dispatch associated with graphics processing units

A computer-implemented method, comprising:

• processing, by a first graphics processing unit, a first data block from a set of data blocks based on a first policy to generate first information for a matrix factorization system;
• dividing, by a second graphics processing unit, a second data block from the set of data blocks into a set of sub-blocks, wherein the second data block comprises disjoint, unrelated features as compared to the first data block;
• processing, by the second graphics processing unit, the set of sub-blocks based on a second policy to generate second information for the matrix factorization system;
• and
• generating, by the central processing unit, a machine learning model for the matrix factorization system based on at least the first information provided by the first graphics processing unit and the second information provided by the second graphics processing unit.

A computer program product for matrix factorization, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a central processing unit to cause the central processing unit to:

• divide a data matrix associated with a matrix factorization system into a set of data blocks;
• assign a first data block from the set of data blocks to a first graphics processing unit;
• assign a second data block from the set of data blocks to a second graphics processing unit, wherein the second data block comprises unrelated, mutually exclusive features as compared to the first data block;
• and
• generate a machine learning model for the matrix factorization system based on at least first information associated with the first data block that is determined by the first graphics processing unit, and the second information associated with the second data block that is determined by the second graphics processing unit.