3 ms·

Could you link a practical algorithm with an exponent lower than 3? (I think of these things https://en.wikipedia.org/wiki/Computational_complexity_of_matrix_mu

Could you link a practical algorithm with an exponent lower than 3? (I think of these things https://en.wikipedia.org/wiki/Computational_complexity_of_matrix_multiplication https://en.wikipedia.org/wiki/Computational_complexity_of_ma... as not being practical, but I'd love to be wrong. )

daniel-cussen 4y ago
https://www.fgemm.com https://www.fgemm.com, coming soon.
gnufx 4y ago
For something Strassen-ish you could look at https://jianyuhuang.com/papers/sc16.pdf https://jianyuhuang.com/papers/sc16.pdf and the GPU implementation https://apps.cs.utexas.edu/apps/sites/default/files/tech_reports/GPUStrassen.pdf https://apps.cs.utexas.edu/apps/sites/default/files/tech_rep...
- FabHK 4y ago
  That's matrix-matrix multiplication. Nobody disputes that Strassen etc. have sub-cubic complexity. What about one of the six decompositions mentioned, as GP claimed?
  - gnufx 4y ago
    I responded to a post about the practicality of MM multiplication methods, though GEMM is quite fundamental.