Numerically Inverting a Matrix and Its Complexity

Reported in interviews at Radix Trading

Problem statement

You are at a whiteboard and asked: given a general $n \times n$ matrix $A$, how would you compute its inverse numerically, and what is the computational complexity?

Describe the approach a real numerical library actually uses, state the asymptotic cost, and mention when you would avoid forming the inverse explicitly and what you would do instead.

Show hint

Which linear system, with many right-hand sides that all share $A$, has the inverse as its solution?

Relevant roles: ML Engineer, Research Scientist and Applied Scientist.

Related free problems