Minimisation methods

K variation:

• conjugate gradients
• search directions corrected for non-orthogonality of localised functions - which becomes more important as localisation range increases