Searching For Fast Matrix Multiplication Algorithms

2017 SIAM Conference on Computational Science and Engineering

Abstract. Matrix multiplication is a powerful tool in computing. However, for the large problems modern data leads to, the traditional algorithm takes too much computation at O(n^3) flops. Strassen provided the first recursive matrix multiplication algorithm in 1969, and many others have improved on the exponent over the following years. We will consider the connection of low rank tensor decompositions to fast matrix multiplication algorithms, and how we are searching for new algorithms that demonstrate the same symmetries as the matrix multiplication tensor.


  • Kathryn Margaret Zuhr Rouse, Wake Forest University, USA,