ajkerr0 / ballnspring Goto Github PK

Maybe we need to enforce certain sparse matrix data types or play around with the routine kwargs.

Our Hessians will be relatively sparse, so we should consider implementing scipy's sparse matrix routines for eigenmode calculation. Maybe issue a warning if our matrix is not 'sparse enough'? We should make an effort to compare the speed and accuracy of the different implementations.

For the genetic algorithms that use this package where larger numbers are desired, the negative values as a result of the arbitrary flipping between hot and cold baths need to be handled. Maybe in the genetic algorithm just take the absolute values of the tcond.