class MAT < $MAT{FLT,VEC,MAT} |
---|
**** | Includes some functions that only work with FLTs Generalizing these functions is possible, but would require definitions of machine epsilon in the numeric classes |
$MAT{_,_,_} | NUMERIC_MAT{_,_} | MAT{_,_} | AREF{_} |
inplace_affine_fit_of(vin,vout:ARRAY{VEC}) |
---|
**** | Fill vin `self' to be the best least squares affine map relating `in' to `out' by: `out[i]=self.affine_act_on(vin[i])'. |
inplace_linear_fit_of(vin,vout:ARRAY{VEC}):MAT |
---|
**** | Fill vin `self' to be the least squares best linear approximation relating `vin' to `vout' by: `out[i]=self.act_on(in[i])'. Return `self'. |
inplace_uniform_random |
---|
**** | Become self's entries uniform in `[0.,1.)' |
inplace_weighted_affine_fit_of(vin,vout:ARRAY{VEC}, wt:ARRAY{FLT}) |
---|
inplace_weighted_linear_fit_of(vin,vout:ARRAY{VEC}, wt:ARRAY{FLT}) |
---|
**** | Fill in `self' to be the least squares best linear approximation relating `vin' to `vout' by: `vout[i]=self.act_on(vin[i])'. `wt[i]' gives the weight which should be given to the ith example. (typically in `[0.,1.]' (`0.' means ignore, `1.' means full weight). |
svd_back_sub(u:MAT, w:VEC, v:MAT, b,x:VEC) |
---|
**** | Solves `a.x=b' for `x' when `a=u.d.v^T' is the svd of `a'. |
svd_in(a:MAT, w:VEC, v:MAT) |
---|
**** | Computes the singular value decomposition of `self = a w v^T'. `a' must be `max(nr,nc)' by `nc', `w' length `nc', `v' is `nc' by `nc'. `Self' is unchanged, `a', `w', `v' are altered. |