Abstract
Matrix multiplication is an essential part of many applications, such as linear algebra, image processing and machine learning. One platform used in such applications is TensorFlow, which is a machine learning library whose structure is based on dataflow programming paradigm. In this work, a method for multiplication of medium-density matrices on multicore CPUs using TensorFlow platform is proposed. This method, called tbt_matmul, utilizes TensorFlow built-in methods tf.matmul and tf.sparse_matmul. By partitioning each input matrix into four smaller sub-matrices, called tiles, and applying an appropriate multiplication method to each pair depending on their density, the proposed method outperforms the built-in methods for matrices of medium density and matrices of significantly uneven distribution of non-zeros.
Highlights
Matrix multiplication is a basis for computation in many areas, such as linear algebra, machine learning and image processing
Since matrix multiplication is often used in many applications, various implementations of matrix multiplication are studied in different environments
It is suggested that, using TensorFlow on multicore CPUs, the multiplication method for sparse matrices could be used, instead of the dense matrix multiplication method, when the density of the input matrices is lower than 70% [13]
Summary
Matrix multiplication is a basis for computation in many areas, such as linear algebra, machine learning and image processing. In these algorithms all matrices are treated as dense matrices Another approach to improve the efficiency of matrix multiplication is based on the sparsity of matrices. We propose an approach to reduce the computation time for medium-density matrix multiplication This approach divides a medium-density matrix into four equal-size submatrices and chooses an appropriate matrix multiplication method for each pair of sub-matrices, based on the density of the sub-matrices. Using this approach, we implement a matrix multiplication function for medium-density matrices on TensorFlow.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
