Approaches to Bounding the Exponent of Matrix Multiplication
summary
Survey describes embedding n×n multiplication into group algebras and into commutative association schemes, using the triple product property to reduce to block-diagonal multiplications; constructs (wreath-product) examples that produce exponents 2.48 and 2.41 but fail to reach 2 because large irreducible dimensions persist. Defines S-rank and proves Omega ≤ (3 Omega_s - 2)/2, so proving Omega_s → 2 would imply Omega → 2. Work produces concrete constructions and numeric bounds but the main conjecture remains open.
AI-written summary · about
chapters (22)
- 0:00Introduction
- 1:11History
- 2:02Outline
- 2:33The matrix multiplication tensor
- 9:08Upper bounds on rank
- 14:55A different approach
- 16:44The general approach
- 21:40The Group Algebra
- 22:25"Nice basis" embedding
- 23:18The embedding
- 25:28How many multiplications?
- 30:15The triangle construction
- 37:25Constructions in linear groups
- 42:06Group algebra approach
- 46:58General semi-simple algebras
- 48:00Weighted vs. unweighted MM
- 52:21Coherent configurations group theory without groups
- 53:33Coherent configs: examples
- 55:18Nice basis conditions
- 55:41Coherent configs vs. groups
- 57:21Proof idea
- 57:47Open problems
