saberbud/ceres-solver
1
0
Fork 0
mirror of https://github.com/ceres-solver/ceres-solver.git synced 2026-08-29 08:34:37 +08:00
Code Issues Packages Projects Releases Wiki Activity

Untabify changes from Jim Roseborough

Browse Source 
Change-Id: Ic640b34ba785669b415acfbeb2c931bea768f985
This commit is contained in:
Sameer Agarwal
2017-04-13 13:17:29 -07:00
parent d04347af43
commit 5ee2b356b3
2 changed files with 35 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/source/analytical_derivatives.rst
+33 -33
View File
@@ -58,22 +58,22 @@ With these derivatives in hand, we can now implement the
virtual ~Rat43Analytic() {} virtual ~Rat43Analytic() {}
virtual bool Evaluate(double const* const* parameters, virtual bool Evaluate(double const* const* parameters,
double* residuals, double* residuals,
double** jacobians) const { double** jacobians) const {
const double b1 = parameters[0][0]; const double b1 = parameters[0][0];
const double b2 = parameters[0][1]; const double b2 = parameters[0][1];
const double b3 = parameters[0][2]; const double b3 = parameters[0][2];
const double b4 = parameters[0][3]; const double b4 = parameters[0][3];
residuals[0] = b1 * pow(1 + exp(b2 - b3 * x_), -1.0 / b4) - y_; residuals[0] = b1 * pow(1 + exp(b2 - b3 * x_), -1.0 / b4) - y_;
if (!jacobians) return true; if (!jacobians) return true;
double* jacobian = jacobians[0]; double* jacobian = jacobians[0];
if (!jacobian) return true; if (!jacobian) return true;
jacobian[0] = pow(1 + exp(b2 - b3 * x_), -1.0 / b4); jacobian[0] = pow(1 + exp(b2 - b3 * x_), -1.0 / b4);
jacobian[1] = -b1 * exp(b2 - b3 * x_) * jacobian[1] = -b1 * exp(b2 - b3 * x_) *
pow(1 + exp(b2 - b3 * x_), -1.0 / b4 - 1) / b4; pow(1 + exp(b2 - b3 * x_), -1.0 / b4 - 1) / b4;
jacobian[2] = x_ * b1 * exp(b2 - b3 * x_) * jacobian[2] = x_ * b1 * exp(b2 - b3 * x_) *
pow(1 + exp(b2 - b3 * x_), -1.0 / b4 - 1) / b4; pow(1 + exp(b2 - b3 * x_), -1.0 / b4 - 1) / b4;
jacobian[3] = b1 * log(1 + exp(b2 - b3 * x_)) * jacobian[3] = b1 * log(1 + exp(b2 - b3 * x_)) *
pow(1 + exp(b2 - b3 * x_), -1.0 / b4) / (b4 * b4); pow(1 + exp(b2 - b3 * x_), -1.0 / b4) / (b4 * b4);
@@ -97,27 +97,27 @@ improve its efficiency, which would give us something like:
virtual ~Rat43AnalyticOptimized() {} virtual ~Rat43AnalyticOptimized() {}
virtual bool Evaluate(double const* const* parameters, virtual bool Evaluate(double const* const* parameters,
double* residuals, double* residuals,
double** jacobians) const { double** jacobians) const {
const double b1 = parameters[0][0]; const double b1 = parameters[0][0];
const double b2 = parameters[0][1]; const double b2 = parameters[0][1];
const double b3 = parameters[0][2]; const double b3 = parameters[0][2];
const double b4 = parameters[0][3]; const double b4 = parameters[0][3];
const double t1 = exp(b2 - b3 * x_); const double t1 = exp(b2 - b3 * x_);
const double t2 = 1 + t1; const double t2 = 1 + t1;
const double t3 = pow(t2, -1.0 / b4); const double t3 = pow(t2, -1.0 / b4);
residuals[0] = b1 * t3 - y_; residuals[0] = b1 * t3 - y_;
if (!jacobians) return true; if (!jacobians) return true;
double* jacobian = jacobians[0]; double* jacobian = jacobians[0];
if (!jacobian) return true; if (!jacobian) return true;
const double t4 = pow(t2, -1.0 / b4 - 1); const double t4 = pow(t2, -1.0 / b4 - 1);
jacobian[0] = t3; jacobian[0] = t3;
jacobian[1] = -b1 * t1 * t4 / b4; jacobian[1] = -b1 * t1 * t4 / b4;
jacobian[2] = -x_ * jacobian[1]; jacobian[2] = -x_ * jacobian[1];
jacobian[3] = b1 * log(t2) * t3 / (b4 * b4); jacobian[3] = b1 * log(t2) * t3 / (b4 * b4);
return true; return true;
} }
private: private:
@@ -182,11 +182,11 @@ When should you use analytical derivatives?
.. rubric:: Footnotes .. rubric:: Footnotes
.. [#f1] The notion of best fit depends on the choice of the objective .. [#f1] The notion of best fit depends on the choice of the objective
function used to measure the quality of fit, which in turn function used to measure the quality of fit, which in turn
depends on the underlying noise process which generated the depends on the underlying noise process which generated the
observations. Minimizing the sum of squared differences is observations. Minimizing the sum of squared differences is
the right thing to do when the noise is `Gaussian the right thing to do when the noise is `Gaussian
<https://en.wikipedia.org/wiki/Normal_distribution>`_. In <https://en.wikipedia.org/wiki/Normal_distribution>`_. In
that case the optimal value of the parameters is the `Maximum that case the optimal value of the parameters is the `Maximum
Likelihood Estimate Likelihood Estimate
<https://en.wikipedia.org/wiki/Maximum_likelihood_estimation>`_. <https://en.wikipedia.org/wiki/Maximum_likelihood_estimation>`_.
docs/source/automatic_derivatives.rst
+2 -2
View File
@@ -39,7 +39,7 @@ implements an automatically differentiated ``CostFunction`` for `Rat43
CostFunction* cost_function = CostFunction* cost_function =
new AutoDiffCostFunction<Rat43CostFunctor, 1, 4>( new AutoDiffCostFunction<Rat43CostFunctor, 1, 4>(
new Rat43CostFunctor(x, y)); new Rat43CostFunctor(x, y));
Notice that compared to numeric differentiation, the only difference Notice that compared to numeric differentiation, the only difference
when defining the functor for use with automatic differentiation is when defining the functor for use with automatic differentiation is
@@ -220,7 +220,7 @@ play. The following code fragment has a simple implementation of a
template <int N> Jet<N> pow(const Jet<N>& f, const Jet<N>& g) { template <int N> Jet<N> pow(const Jet<N>& f, const Jet<N>& g) {
return Jet<N>(pow(f.a, g.a), return Jet<N>(pow(f.a, g.a),
g.a * pow(f.a, g.a - 1.0) * f.v + g.a * pow(f.a, g.a - 1.0) * f.v +
pow(f.a, g.a) * log(f.a); * g.v); pow(f.a, g.a) * log(f.a); * g.v);
} }