This removes the determinant calculation inside the CG method. Speeds up "slow" 2x2 solvers. All the determinant is likely to catch is singular matrices - these will likely only occur when a gain has no data points contributing to it. Catching diverging gains will need to be more sophisticated and likely examine the behaviour of the updates themselves.
This removes the determinant calculation inside the CG method. Speeds up "slow" 2x2 solvers. All the determinant is likely to catch is singular matrices - these will likely only occur when a gain has no data points contributing to it. Catching diverging gains will need to be more sophisticated and likely examine the behaviour of the updates themselves.