Model fit

Rather than fitting a model

\[\min_z l(A z + b) + r(z),\]

via a specific choice of a loss function, \(l\), and regularization function, \(r\), such as for a support vector machine or logistic regression, the following routines accept function wrappers for the proximal maps said functions and execute an Alternating Direction Method of Multipliers.

Note

While sometimes functional, this implementation is still very much a work-in-progress.

C++ API

Int ModelFit(std::function<void(Matrix<Real>&, Real)> lossProx, std::function<void(Matrix<Real>&, Real)> regProx, const Matrix<Real> &A, const Matrix<Real> &b, Matrix<Real> &w, Real rho, Int maxIter = 1000, bool inv = true, bool progress = true, )

Int ModelFit(std::function<void(DistMatrix<Real>&, Real)> lossProx, std::function<void(DistMatrix<Real>&, Real)> regProx, const AbstractDistMatrix<Real> &A, const AbstractDistMatrix<Real> &b, AbstractDistMatrix<Real> &w, Real rho, Int maxIter = 1000, bool inv = true, bool progress = true, )

C API

ElError ElModelFit_s(void (*lossProx)(ElMatrix_s,float), void (*regProx)(ElMatrix_s,float), ElConstMatrix_s A, ElConstMatrix_s b, ElMatrix_s w, float rho, ElInt* numIts)

ElError ElModelFit_d(void (*lossProx)(ElMatrix_d,double), void (*regProx)(ElMatrix_d,double), ElConstMatrix_d A, ElConstMatrix_d b, ElMatrix_d w, double rho, ElInt* numIts)

ElError ElModelFitDist_s(void (*lossProx)(ElDistMatrix_s,float), void (*regProx)(ElDistMatrix_s,float), ElConstDistMatrix_s A, ElConstDistMatrix_s b, ElDistMatrix_s w, float rho, ElInt* numIts)

ElError ElModelFitDist_d(void (*lossProx)(ElDistMatrix_d,double), void (*regProx)(ElDistMatrix_d,double), ElConstDistMatrix_d A, ElConstDistMatrix_d b, ElDistMatrix_d w, double rho, ElInt* numIts)