{"id":"arxiv:2402.01011","kind":"preprint","title":"Ruling Out Low-rank Matrix Multiplication Tensor Decompositions with Symmetries via SAT","authors":["Jason Yang"],"author_count":1,"year":2024,"published":"2024-02-01","venue":null,"cited_by":null,"doi":null,"arxiv_id":"2402.01011","nct_id":null,"source_url":"https://arxiv.org/abs/2402.01011","archive_url":"https://space-kitty.com/papers/arxiv:2402.01011","updated":"2024-02-01","pmid":null,"pmcid":null,"openalex_id":null,"pdf_url":"https://arxiv.org/pdf/2402.01011","abstract":"We analyze rank decompositions of the $3\\times 3$ matrix multiplication tensor over $\\mathbb{Z}/2\\mathbb{Z}$. We restrict our attention to decompositions of rank $\\le 21$, as only those decompositions will yield an asymptotically faster algorithm for matrix multiplication than Strassen's algorithm. To reduce search space, we also require decompositions to have certain symmetries. Using Boolean SAT solvers, we show that under certain symmetries, such decompositions do not exist.","abstract_license":"CC0-1.0 (arXiv descriptive metadata)","license":null,"is_open_access":false,"retracted":false,"peer_reviewed":false,"categories":["cs.CC"],"details":null,"provenance":"arxiv","attribution":"Abstract: arXiv descriptive metadata (CC0). Thank you to arXiv for use of its open access interoperability. This service was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.","topics":[],"cited_in_dossiers":[]}