Changed DynamicAutoDiffCostFunction to handle multiple derivative
sections as opposed to just a single contiguous block.
In the previous implementation it was assumed that non-constant
parameters occur in a single contiguous block so that constant
parameters could NOT lie between non-constant parameters. Previously,
start_derivative_section was first set as soon as the first
non-constant parameter block (marked by jacobians[i] != NULL) was
encountered. After this, entries in input_jets[parameter_cursor].v were
accessed with `parameter_cursor - start_derivative_section`. For
contiguous non-constant parameter blocks this is fine, but if constant
parameter blocks fall between then this indexing is incorrect because
`parameter_cursor - start_derivative_section` can go out of bounds.
For a concrete example, take a cost function with three parameter
blocks, each of size 1 and with the center block fixed. Assume that
Stride=1 so that two passes are required. On the first pass
start_derivative_section=0, and the first variable block is handled
correctly. At the end of the first pass end_derivative_section=1, so
for the second pass start_derivative_section=1. Now comes the problem.
When parameter_cursor=1, parameter_cursor >= start_derivative_section
so jacobian[1] is checked to be NULL. Since it is NULL (second
parameter block is constant) then nothing is done and
active_parameter_count is NOT incremented. Next, when
parameter_cursor=2, parameter_cursor >= start_derivative_section and
jacobian[2] is checked. Since it is not NULL then
input_jets[parameter_cursor].v[parameter_cursor -
start_derivative_section] is set to 1.0, BUT parameter_cursor -
start_derivative_section = 2 - 1 = 1 which is out of bounds
(input_jets[parameter_cursor].v is only of size Stride=1).
The proposed solution records the start of each contiguous block of
non-constant parameters and indexing into
input_jets[parameter_cursor].v is independent of parameter_cursor.
Change-Id: I388ab6a0bafa35d317491135ec6fe980453ff888
This support was broken due to the TrustRegionMinimizer refactoring.
It is now enabled again, with the responsibilty for dumping the
problem shifted to the individual TrustRegionStrategy.
There is however one wrinkle, which is perhaps an indication of
poor design to start with. The LinearLeastSquaresProblemProto
carries in it num_eliminate_blocks, something which does not
exist anymore. More importantly, the TrustRegionStrategy does not
have access to this quantity anymore.
Dealing with this will be the subject of a future change.
Change-Id: I358adf6a2e386f4940b617bf950d6c7e87d2635d
1. Multithread the inversion of J'J.
2. Simplify the dense rank truncation loop.
3. Minor correction to building documentation.
Change-Id: Ide932811c0f28dc6c253809339fb2caa083865b5
1. Further tightening of the Covariance documentation.
2. Documented minimizer progress output.
3. Lint cleanup from William Rucklidge.
4. Updated version history.
Change-Id: I8bc28484675d4edf89a7c050b6379dbac6c39e91
1. Sparse covariance estimation now uses cholmod_rcond to
detect singular Jacobians.
2. Dense covariance estimation now uses relative magnitude
of singular/eigen values to compute the pseudoinverse.
3. Truncation logic is now unified with Solver::Options::null_space_rank.
Change-Id: I095bd737510c836b4251255926190a7f31d64bce
Also
1. Remove an inadvertent LOG(INFO) from trust_region_minimizer.cc
2. Refactor some of the code in FullReport to reduce duplication
across line search and trust region minimizers.
3. Consistent capitalization.
Change-Id: I9078b1704efab23d2858530636f524e60c7d9016
This allows CovarianceImpl to be forward declared without
scoped_ptr freaking out.
Thanks to Nima Keivan for reporting this.
Change-Id: Icd5aa766b3aab70246055225231a4b971c6b7b90
Add a Covariance object to the API.
Given a Problem object and a set of parameter block pairs the
Covariance object computes a sparse covariance matrix corresponding
to those block pairs and provides random access to them.
Constant parameter blocks and parameter blocks with local parameterizations
are correctly handled.
Sparse and dense implementations are provided. With the dense implementation
rank deficient Jacobians can also be handled.
Parts of the code are threaded using OpenMP if available.
Change-Id: I5b49583b3d79579df3e0f334c22567acb23ed4ad
This introduces a simple C API for a subset of Ceres. This opens the door to
using languages like Python to call Ceres, since it is much easier to bind to C
than it is to bind to C++. It will mean giving up the native Ceres autodiff.
The implementation in this patch does not attempt to do everything but is only
just enough to get started. Subsequent patches will increase the surface area
of Ceres that is covered by the C API.
Change-Id: Ic51804bac6865e1a2e476553248aabc91dff3409
1. Bug fix in NumericDiffCostFunction (Thanks to Nicolas Brodu).
2. Minor documentation update in solver.h
3. Version history update.
4. Bump the version and ABI version.
Change-Id: I951574ddd0b2c4c03b9c79ff33eb9bea549071e7
Sparse Cholesky factorization algorithms use a fill-reducing
ordering to permute the columns of the Jacobian matrix. There
are two ways of doing this.
1. Compute the Jacobian matrix in some order and then have the
factorization algorithm permute the columns of the Jacobian.
2. Compute the Jacobian with its columns already permuted.
The first option incurs a significant memory penalty. The
factorization algorithm has to make a copy of the permuted
Jacobian matrix.
Starting with this change Ceres pre-permutes the columns of the
Jacobian matrix and generally speaking, there is no performance
penalty for doing so.
In some rare cases, it is worth using a more complicated
reordering algorithm which has slightly better runtime
performance at the expense of an extra copy of the Jacobian
matrix. Setting Solver::Options::use_postordering to true
enables this tradeoff.
This change also removes Solver::Options::use_block_amd
as an option. All matrices are ordered using their block
structure. The ability to order them by their scalar
sparsity structure has been removed.
Here is what performance on looks like on some BAL problems.
Memory
======
HEAD pre-ordering
16-22106 137957376.0 113516544.0
49-7776 56688640.0 46628864.0
245-198739 1718005760.0 1383550976.0
257-65132 387715072.0 319512576.0
356-226730 2014826496.0 1626087424.0
744-543562 4903358464.0 3957878784.0
1024-110968 968626176.0 822071296.0
Time
====
HEAD pre-ordering
16-22106 3.8 3.7
49-7776 1.9 1.8
245-198739 82.6 81.9
257-65132 14.0 13.4
356-226730 98.8 95.8
744-543562 325.2 301.6
1024-110968 42.1 37.1
Change-Id: I6b2e25f3fed7310f88905386a7898ac94d37467e
This class is used to create local parameterization
with Jacobians computed via automatic differentiation.
To get an auto differentiated local parameterization,
class with a templated operator() (a functor) that
computes
plus_delta = Plus(x, delta);
shall be defined.
Then given such functor, the auto differentiated local
parameterization can be constructed as
LocalParameterization* local_parameterization =
new AutoDiffLocalParameterization<PlusFunctor, 4, 3>;
| |
Global Size ---------------+ |
Local Size -------------------+
See autodiff_local_parameterization.h for more information
and usage example.
Initial implementation by Keir Mierle, finished by self
and integrated into Ceres and covered with unit tests
by Sameer Agarwal.
Change-Id: I1b3e48ae89f81e0cf1f51416c5696e18223f4b21
1. Introduce new typdefs in eigen.h to allow for column
major matrices.
2. Clean up old unused typedefs, and the aligned typedefs
since they do not actually add any real performance.
3. Made eigen.h conform to the google style guide by removing
the using directives. They were polluting the ceres namespace.
4. Made the template specialization generator work again.
Change-Id: Ic2268c784534b737ebd6e1a043e2a327adaeca37
1. Add the ability to evaluate the problem without loss function.
2. Remove static Evaluator::Evaluate
3. Refactor the common code from problem_test.cc and
evaluator_test.cc into evaluator_test_utils.cc
Change-Id: I1aa841580afe91d288fbb65288b0ffdd1e43e827
Here is an example report, obtained by running:
bin/Debug/bundle_adjuster \
--input=../ceres-solver/data/problem-16-22106-pre.txt \
--linear_solver=iterative_schur \
--num_iterations=1 \
--alsologtostderr \
--use_local_parameterization \
--use_quaternions
Note that effective parameters is less than parameters by 16, which is the
number of cameras. In this case the local parameterization has a 3 dimensional
tangent space for the 4-dimensional quaternions.
Ceres Solver Report
-------------------
Original Reduced
Parameter blocks 22138 22138
Parameters 66478 66478
Effective parameters 66462 66462
Residual blocks 83718 83718
Residual 167436 167436
Minimizer TRUST_REGION
Trust Region Strategy LEVENBERG_MARQUARDT
Given Used
Linear solver ITERATIVE_SCHUR ITERATIVE_SCHUR
Preconditioner JACOBI JACOBI
Threads: 1 1
Linear solver threads 1 1
Linear solver ordering AUTOMATIC 22106, 32
Cost:
Initial 4.185660e+06
Final 7.221647e+04
Change 4.113443e+06
Number of iterations:
Successful 1
Unsuccessful 0
Total 1
Time (in seconds):
Preprocessor 0.697
Residual Evaluations 0.063
Jacobian Evaluations 27.608
Linear Solver 13.360
Minimizer 43.973
Postprocessor 0.004
Total 44.756
Termination: NO_CONVERGENCE
Change-Id: I6b6b8ac24f71bd187e67d95651290917642be74f
This patch introduces a matrix wrapper (MatrixAdapter) that allows to
transparently pass pointers to row-major or column-major matrices
to the conversion functions.
Change-Id: I7f1683a8722088cffcc542f593ce7eb46fca109b
This adds support for removing parameter and residual blocks.
There are two modes of operation: in the first, removals of
paremeter blocks are expensive, since each remove requires
scanning all residual blocks to find ones that depend on the
removed parameter. In the other, extra memory is sacrificed to
maintain a list of the residuals a parameter block depends on,
removing the need to scan. In both cases, removing residual blocks
is fast.
As a caveat, any removals destroys the ordering of the parameters,
so the residuals or jacobian returned from Solver::Solve() is
meaningless. There is some debate on the best way to handle this;
the details remain for a future change.
This also adds some overhead, even in the case that fast removals
are not requested:
- 1 int32 to each residual, to track its position in the program.
- 1 pointer to each parameter, to store the dependent residuals.
Change-Id: I71dcac8656679329a15ee7fc12c0df07030c12af
1. Added a Preconditioner interface.
2. SCHUR_JACOBI is now its own class and is independent of
SuiteSparse.
Change-Id: Id912ab19cf3736e61d1b90ddaf5bfba33e877ec4
1. Add an ExecutionSummary object to record execution
information about Ceres objects.
2. Add an EventLogger object to log events in a function call.
3. Add a ScopedExecutionTimer object to log times in ExecutionSummary.
4. Instrument ProgramEvaluator and all the linear solvers
to report their timing statistics.
5. Connect the timing statistics to Summary::FullReport.
6. Add high precision timer on unix systems using
gettimeofday() call.
7. Various minor clean ups all around.
Change-Id: I5e09804b730b09535484124be7dbc1c58eccd1d4
A wrapper class that takes a variadic functor evaluating a
function, numerically differentiates it and makes it available as a
templated functor so that it can be easily used as part of Ceres'
automatic differentiation framework.
The tests for NumericDiffCostFunction and NumericDiffFunctor have
a lot of stuff that is common, so refactor them to reduce code.
Change-Id: I83b01e58b05e575fb2530d15cbd611928298646a
Richard Stebbing