/sci/ - Science & Math

>>11889375
Unrelated serious questions:
I'm doing control and identification on real world sensor data - I take long sequences of data, tensor product the vector to itself to get a matrix, and then I try to take the inverse of that matrix.

because its real world data, sometimes the determinant is big, sometimes its small and sometimes its very small - enough to fuck up any attempt at an inverse matrix. yet there's nothing inherently singular about the process generating the matrix.

Is there some technique to get some approx equivalent version of an inverse from an almost singular matrix? Something involving approximation/ linear nonlinear programming?