OpenGrok
Home
Sort by relevance
Sort by last modified time
Full Search
Definition
Symbol
File Path
History
|
|
Help
Searched
full:hessian
(Results
1 - 18
of
18
) sorted by null
/external/ceres-solver/internal/ceres/
low_rank_inverse_hessian.h
32
//
Hessian
, using the LBFGS algorithm
44
//
Hessian
using the limited memory variant of the
46
// approximating the
Hessian
.
62
// num_parameters is the row/column size of the
Hessian
.
63
// max_num_corrections is the rank of the
Hessian
approximation.
65
// inverse
Hessian
used during Right/LeftMultiply() is scaled by
66
// the approximate eigenvalue of the true inverse
Hessian
at the
77
// domain of
Hessian
, and delta_gradient is the change in the
line_search_direction.cc
117
<< "Ceres bug: NextDirection() called on L-BFGS after inverse
Hessian
"
130
LOG(WARNING) << "Numerical failure in L-BFGS update: inverse
Hessian
"
156
<< " parameters, this will allocate a dense approximate inverse
Hessian
"
171
<< "Ceres bug: NextDirection() called on BFGS after inverse
Hessian
"
183
// Update dense inverse
Hessian
approximation.
186
// Rescale the initial inverse
Hessian
approximation (H_0) to be
188
// inverse
Hessian
at the start point. As shown in [1]:
198
// of the true initial
Hessian
(not the inverse) respectively. Thus,
199
// \gamma is an approximate eigenvalue of the true inverse
Hessian
, and
201
// similar scale to the true inverse
Hessian
. This technique is widel
[
all
...]
low_rank_inverse_hessian.cc
105
// Rescale the initial inverse
Hessian
approximation (H_0) to be iteratively
106
// updated so that it is of similar 'size' to the true inverse
Hessian
along
117
// the true
Hessian
(not the inverse) along the most recent search direction
119
// inverse
Hessian
, and choosing: H_0 = I * \gamma will yield a starting
120
// point that has a similar scale to the true inverse
Hessian
. This
corrector.cc
61
//
Hessian
gets both the scaling and the rank-1 curvature
75
// newton
hessian
goes from being a full rank correction to a rank
76
// deficient correction making the inversion of the
Hessian
fraught
corrector_test.cc
163
// Corrected
hessian
and gradient implied by the modified jacobian
202
// Corrected gradient and
hessian
.
231
// Corrected
hessian
and gradient implied by the modified jacobian
263
// Corrected gradient and
hessian
.
coordinate_descent_minimizer.h
47
// (non-exhaustively) the
Hessian
matrix into independent sets,
parameter_block_ordering.h
75
// structure reflects the sparsity structure of the
Hessian
. Each
schur_ordering.h
65
// structure reflects the sparsity structure of the
Hessian
. Each
schur_eliminator_impl.h
420
// and the off diagonal blocks in the Guass Newton
Hessian
.
solver_impl.cc
[
all
...]
/external/opencv/cv/src/
cvsurf.cpp
10
* 2.A comparision with original libSurf.so shows that the
hessian
detector is not a 100% match to their implementation;
122
/*
hessian
detector */
144
float*
hessian
= hessians[k]->data.fl;
local
148
hessian
[i] =
hessian
[hessian_cols*hessian_rows-1-i] =
151
hessian
+= (SIZE0/2)*(hessian_cols + 1);
158
trace += hessian_cols,
hessian
+= hessian_cols )
162
hessian
[-j-1] =
hessian
[hessian_cols - SIZE0 + j] =
177
hessian
[j] = (float)(dx*dy - dxy*dxy)
196
const float*
hessian
=
hessian
s[k]->data.fl + i*
hessian
_cols;
local
218
const float*
hessian
=
hessian
s[z]->data.fl + (j*scale+scaleCache[z]\/2)\/scaleCache[z]-1 +
local
[
all
...]
/external/ceres-solver/include/ceres/
types.h
102
// Block diagonal of the Gauss-Newton
Hessian
.
178
// algorithms that approximate the
Hessian
matrix by iteratively refining
184
// equation. The requirement that the
Hessian
approximation be positive
187
// approximate
Hessian
by imposing the additional constraints that the
195
// maintains a full, dense approximation to the (inverse)
Hessian
, L-BFGS
199
// full dense inverse
Hessian
approximation. This is particularly important
solver.h
161
// The LBFGS
hessian
approximation is a low rank approximation to
162
// the inverse of the
Hessian
matrix. The rank of the
171
// 2. The
Hessian
approximation is constrained to be positive
188
// the initial inverse
Hessian
approximation is taken to be the Identity.
190
// chosen to approximate an eigenvalue of the true inverse
Hessian
can
553
//
Hessian
matrix's sparsity structure in a collection of
[
all
...]
/external/webrtc/src/modules/audio_coding/codecs/isac/main/source/
pitch_estimator.c
530
/* gradient and approximate
Hessian
(lower triangle) for minimizing the filter's output power */
546
/* add gradient and
Hessian
(lower triangle) for dampening fast gain changes */
559
/* add gradient and
Hessian
for dampening gain */
570
/* compute Cholesky factorization of
Hessian
/external/ceres-solver/docs/source/
solving.rst
300
Hessian
matrix's sparsity structure into a collection of
362
Here :math:`H(x)` is some approximation to the
Hessian
of the
391
Hessian
is maintained and used to compute a quasi-Newton step
397
inverse
Hessian
used to compute a quasi-Newton step [Nocedal]_,
[
all
...]
modeling.rst
[
all
...]
/external/opencv/cv/include/
cv.h
1084
float
hessian
;
member in struct:CvSURFPoint
[
all
...]
/external/ceres-solver/examples/
nist.cc
107
"Rank of L-BFGS inverse
Hessian
approximation in line search.");
Completed in 484 milliseconds