Next: Zero finders, Previous: Scalar optimization, Up: Top [Index]
Model functions whose parameters are to be optimized may return a vector or array of values. Either these or their differences to some constant values (curve fitting) can be minimized in some sense, often, but not necessarily, by minimizing the sum of their squares. It is usually preferable to use optimizers designed for residual optimization for this purpose. These can exploit information contained in the individual elements of the returned array, which would not be possible if the user calculated a norm (e.g. sum of squares) of the elements and performed a scalar optimization.
Optimization frontends | ||
|---|---|---|
| • nonlin_residmin: | The standard interface for non-linear residual minimization. | |
| • nonlin_curvefit: | A convenience interface, curve fitting. | |
Optimization backends | ||
| • lm_svd_feasible: | L/M algorithm with SVD, constraints met throughout optimization. | |
Statistics frontends | ||
| • residmin_stat: | Statistics for residual minimization. | |
| • curvefit_stat: | Statistics for curve fitting. | |
Statistics backends | ||
| • wls: | Statistics for weighted least squares. | |
Standalone functions | ||
| • lsqlin: | Linear least squares with linear constraints. | |
| • leasqr: | An older function for curve fitting. | |
| • pronyfit: | Prony’s method for non-linear exponential fitting. | |
| • polyfitinf: | Function polyfitinf for polynomial fitting. | |
| • wpolyfit: | Polynomial fitting suitable for polyconf. | |
| • polyconf: | Confidence and prediction intervals for polynomial fitting. | |
| • LinearRegression: | Function LinearRegression. | |
| • wsolve: | Another linear solver. | |
Next: Zero finders, Previous: Scalar optimization, Up: Top [Index]