a linear-time method for computing the Hessian log determinant would yield an O(n^2) matrix multiplication algorithm.
Are there bounds (or “bounds” like O(n^2) matmul being unlikely to come soon or to have galaxy scale constant terms) on approximate matrix multiplication? My intuition tells me that for, say, sparse matrices, you may get a large speedup.
Are there bounds (or “bounds” like O(n^2) matmul being unlikely to come soon or to have galaxy scale constant terms) on approximate matrix multiplication? My intuition tells me that for, say, sparse matrices, you may get a large speedup.