Corrected typo in comment.

The inverted JtJ does not multiply itself, but rather Jte. Correct this
in the comment.
This commit is contained in:
Olexa Bilaniuk 2015-02-04 12:12:14 -05:00
parent 2609e77af7
commit b90800f0c7

View File

@ -1583,7 +1583,7 @@ static inline void sacRefine(RHO_HEST_REFC* p){
int i;
float S, newS; /* Sum of squared errors */
float gain; /* Gain-parameter. */
float L = 0.01f;/* Lambda of LevMarq */
float L = 100.0f;/* Lambda of LevMarq */
float dH[8], newH[8];
/**
@ -1614,7 +1614,7 @@ static inline void sacRefine(RHO_HEST_REFC* p){
* The system above is solved by Cholesky decomposition of a
* sufficently-damped JtJ into a lower-triangular matrix (and its
* transpose), whose inverse is then computed. This inverse (and its
* transpose) then multiply JtJ in order to find dH.
* transpose) then multiply Jte in order to find dH.
*/
while(!sacChol8x8Damped(p->lm.JtJ, L, p->lm.tmp1)){