Polyhedral Homotopy Continuation Method for solving sparse polynomial system, optimized by only tracing real zeros
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

1029 lines
34 KiB

// This file is part of Bertini 2.0.
//
// trackers/explicit_predictors.hpp is free software: you can redistribute it
// and/or modify it under the terms of the GNU General Public License as
// published by the Free Software Foundation, either version 3 of the License,
// or (at your option) any later version.
//
// trackers/explicit_predictors.hpp is distributed in the hope that it will be
// useful, but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License for more details.
//
// You should have received a copy of the GNU General Public License
// along with trackers/explicit_predictors.hpp. If not, see
// <http://www.gnu.org/licenses/>.
//
// copyright 2015
// James B. Collins
// West Texas A&M University
// Department of Mathematics
// Spring 2016
//
// silviana amethyst, university of wisconsin-eau claire
/**
\file explicit_predictors.hpp
\brief Contains a base class for all ODE predictors.
*/
#ifndef BERTINI_EXPLICIT_PREDICTORS_HPP
#define BERTINI_EXPLICIT_PREDICTORS_HPP
#include <Eigen/LU>
#include <boost/type_index.hpp>
#include "eigen_extensions.hpp"
#include "mpfr_extensions.hpp"
#include "system/system.hpp"
#include "trackers/amp_criteria.hpp"
namespace bertini {
namespace tracking {
namespace predict {
// Constant,
// Euler,
// Heun,
// RK4,
// HeunEuler,
// RKNorsett34,
// RKF45,
// RKCashKarp45,
// RKDormandPrince56,
// RKVerner67
/**
\brief Get the Bertini2 default predictor.
Currently set to RKF45.
\return The default predictor method to use.
*/
inline Predictor DefaultPredictor() { return Predictor::RKF45; }
/**
\brief The order of the predictor.
The order of the error estimate is this plus one.
\return The order of the predictor method.
\param predictor_choice The predictor method to query.
*/
inline unsigned Order(Predictor predictor_choice) {
switch (predictor_choice) {
case (Predictor::Constant):
return 0;
case (Predictor::Euler):
return 1;
case (Predictor::Heun):
return 2;
case (Predictor::HeunEuler):
return 2;
case (Predictor::RKNorsett34):
return 3;
case (Predictor::RK4):
return 4;
case (Predictor::RKF45):
return 4;
case (Predictor::RKCashKarp45):
return 4;
case (Predictor::RKDormandPrince56):
return 5;
case (Predictor::RKVerner67):
return 6;
default: {
throw std::runtime_error("incompatible predictor choice in Order");
}
}
}
/**
\brief Ask whether a predictor method provides an error estimate.
\return Yes or no, does it or does it not.
\param predictor_choice The predictor method to query.
*/
inline bool HasErrorEstimate(Predictor predictor_choice) {
switch (predictor_choice) {
case (Predictor::Constant):
return false;
case (Predictor::Euler):
return false;
case (Predictor::Heun):
return false;
case (Predictor::HeunEuler):
return true;
case (Predictor::RKNorsett34):
return true;
case (Predictor::RK4):
return false;
case (Predictor::RKF45):
return true;
case (Predictor::RKCashKarp45):
return true;
case (Predictor::RKDormandPrince56):
return true;
case (Predictor::RKVerner67):
return true;
default: {
throw std::runtime_error(
"incompatible predictor choice in HasErrorEstimate");
}
}
}
namespace {
template <typename T>
struct LUSelector {};
template <>
struct LUSelector<dbl> {
template <typename N>
static Eigen::PartialPivLU<Mat<dbl>>& Run(N& n) {
return n.GetLU_d();
}
};
template <>
struct LUSelector<mpfr_complex> {
template <typename N>
static Eigen::PartialPivLU<Mat<mpfr_complex>>& Run(N& n) {
return n.GetLU_mp();
}
};
} // namespace
/**
\class ExplicitRKPredictor
\brief A class which stores all the explicit single-step multi-stage ODE
predictor methods.
## Purpose
Stores all the information needed to implement the predictor method
- Butcher Table
- Number of Stages
- Order of the method.
Also stores information computed during implementation of the method.
## Use
Each predictor method is stored as a static Butcher table. To perform a
predict step, you must instantiate an object with a particular predictor method
and call Predict:
\code
ExplicitRKPredictors<Complex,Real> euler(Predictor::Euler, sys)
success_code = euler.Predict( ... )
\endcode
*/
class ExplicitRKPredictor {
friend LUSelector<dbl>;
friend LUSelector<mpfr_complex>;
public:
/**
\brief Construct a predictor to work on a system.
\param S the system the predictor will be predicting on.
*/
ExplicitRKPredictor(const System& S)
: current_precision_(DefaultPrecision()), s_(0) {
ChangeSystem(S);
PredictorMethod(DefaultPredictor());
}
/**
\brief Constructor for a particular predictor method
\param method The predictor method to be implemented.
\param S the system to be predicting on.
*/
ExplicitRKPredictor(Predictor method, const System& S)
: current_precision_(DefaultPrecision()), s_(0) {
ChangeSystem(S);
PredictorMethod(method);
}
/**
/brief Sets the local variables to correspond to a particular predictor
method
\param method Enum class that determines the predictor method
*/
void PredictorMethod(Predictor method) {
predictor_ = method;
p_ = predict::Order(method);
switch (method) {
case Predictor::Constant: {
s_ = 1;
Mat<double>& arefd = std::get<Mat<double>>(a_);
Vec<double>& brefd = std::get<Vec<double>>(b_);
Vec<double>& crefd = std::get<Vec<double>>(c_);
crefd.resize(s_);
crefd(0) = 0;
arefd.resize(s_, s_);
arefd(0, 0) = 0;
brefd.resize(s_);
brefd(0) = 0;
Mat<mpfr_float>& arefmp = std::get<Mat<mpfr_float>>(a_);
Vec<mpfr_float>& brefmp = std::get<Vec<mpfr_float>>(b_);
Vec<mpfr_float>& crefmp = std::get<Vec<mpfr_float>>(c_);
crefmp.resize(s_);
crefmp(0) = 0;
arefmp.resize(s_, s_);
arefmp(0, 0) = 0;
brefmp.resize(s_);
brefmp(0) = 0;
uses_embedded_ = false;
break;
}
case Predictor::Euler: {
s_ = 1;
Mat<double>& arefd = std::get<Mat<double>>(a_);
Vec<double>& brefd = std::get<Vec<double>>(b_);
Vec<double>& crefd = std::get<Vec<double>>(c_);
crefd.resize(s_);
crefd(0) = static_cast<double>(cEuler_(0));
arefd.resize(s_, s_);
arefd(0, 0) = static_cast<double>(aEuler_(0, 0));
brefd.resize(s_);
brefd(0) = static_cast<double>(bEuler_(0));
Mat<mpfr_float>& arefmp = std::get<Mat<mpfr_float>>(a_);
Vec<mpfr_float>& brefmp = std::get<Vec<mpfr_float>>(b_);
Vec<mpfr_float>& crefmp = std::get<Vec<mpfr_float>>(c_);
crefmp.resize(s_);
crefmp(0) = static_cast<mpfr_float>(cEuler_(0));
arefmp.resize(s_, s_);
arefmp(0, 0) = static_cast<mpfr_float>(aEuler_(0, 0));
brefmp.resize(s_);
brefmp(0) = static_cast<mpfr_float>(bEuler_(0));
uses_embedded_ = false;
break;
}
case Predictor::HeunEuler: {
s_ = 2;
FillButcherTable<double>(s_, aHeunEuler_, bHeunEuler_,
b_minus_bstarHeunEuler_, cHeunEuler_);
FillButcherTable<mpfr_float>(s_, aHeunEuler_, bHeunEuler_,
b_minus_bstarHeunEuler_, cHeunEuler_);
break;
}
case Predictor::RK4: {
s_ = 4;
FillButcherTable<double>(s_, aRK4_, bRK4_, cRK4_);
FillButcherTable<mpfr_float>(s_, aRK4_, bRK4_, cRK4_);
break;
}
case Predictor::RKF45: {
s_ = 6;
FillButcherTable<double>(s_, aRKF45_, bRKF45_, b_minus_bstarRKF45_,
cRKF45_);
FillButcherTable<mpfr_float>(s_, aRKF45_, bRKF45_, b_minus_bstarRKF45_,
cRKF45_);
break;
}
case Predictor::RKCashKarp45: {
s_ = 6;
FillButcherTable<double>(s_, aRKCK45_, bRKCK45_, b_minus_bstarRKCK45_,
cRKCK45_);
FillButcherTable<mpfr_float>(s_, aRKCK45_, bRKCK45_,
b_minus_bstarRKCK45_, cRKCK45_);
break;
}
case Predictor::RKDormandPrince56: {
s_ = 8;
FillButcherTable<double>(s_, aRKDP56_, bRKDP56_, b_minus_bstarRKDP56_,
cRKDP56_);
FillButcherTable<mpfr_float>(s_, aRKDP56_, bRKDP56_,
b_minus_bstarRKDP56_, cRKDP56_);
break;
}
case Predictor::RKVerner67: {
s_ = 10;
FillButcherTable<double>(s_, aRKV67_, bRKV67_, b_minus_bstarRKV67_,
cRKV67_);
FillButcherTable<mpfr_float>(s_, aRKV67_, bRKV67_, b_minus_bstarRKV67_,
cRKV67_);
break;
}
default: {
throw std::runtime_error(
"incompatible predictor choice in ExplicitPredict");
}
}
ResizeK();
}; // re: PredictorMethod
/**
\brief Change the system (number of total functions) that the predictor uses.
\param S New system to switch to.
*/
void ChangeSystem(const System& S) {
numTotalFunctions_ = S.NumTotalFunctions();
numVariables_ = S.NumVariables();
// you cannot set K_ here, because s_ may not have been set
std::get<Mat<dbl>>(dh_dx_0_).resize(numTotalFunctions_, numVariables_);
std::get<Mat<mpfr_complex>>(dh_dx_0_).resize(numTotalFunctions_,
numVariables_);
std::get<Mat<dbl>>(dh_dx_temp_).resize(numTotalFunctions_, numVariables_);
std::get<Mat<mpfr_complex>>(dh_dx_temp_)
.resize(numTotalFunctions_, numVariables_);
std::get<Vec<dbl>>(dh_dt_temp_).resize(numTotalFunctions_);
std::get<Vec<mpfr_complex>>(dh_dt_temp_).resize(numTotalFunctions_);
ResizeK();
}
void ResizeK() {
std::get<Mat<dbl>>(K_).resize(numTotalFunctions_, s_);
std::get<Mat<mpfr_complex>>(K_).resize(numTotalFunctions_, s_);
}
/**
\brief get the current precision of the predictor
*/
unsigned precision() const { return current_precision_; }
/**
/brief Change the precision of the predictor variables and reassign the
Butcher table variables.
\param new_precision The new precision.
*/
void ChangePrecision(unsigned new_precision) {
Precision(std::get<Mat<mpfr_complex>>(K_), new_precision);
Precision(std::get<Vec<mpfr_complex>>(dh_dt_temp_), new_precision);
Precision(std::get<Mat<mpfr_complex>>(dh_dx_0_), new_precision);
Precision(std::get<Mat<mpfr_complex>>(dh_dx_temp_), new_precision);
Precision(std::get<Mat<mpfr_float>>(a_), new_precision);
Precision(std::get<Vec<mpfr_float>>(b_), new_precision);
Precision(std::get<Vec<mpfr_float>>(b_minus_bstar_), new_precision);
Precision(std::get<Vec<mpfr_float>>(c_), new_precision);
PredictorMethod(predictor_);
current_precision_ = new_precision;
PrecisionSanityCheck();
}
void PrecisionSanityCheck() const {
assert(current_precision_ == DefaultPrecision());
Vec<mpfr_complex>& dhdttemp = std::get<Vec<mpfr_complex>>(dh_dt_temp_);
Mat<mpfr_complex>& dhdx0 = std::get<Mat<mpfr_complex>>(dh_dx_0_);
Mat<mpfr_complex>& dhdxtemp = std::get<Mat<mpfr_complex>>(dh_dx_temp_);
Mat<mpfr_float>& a = std::get<Mat<mpfr_float>>(a_);
Vec<mpfr_float>& b = std::get<Vec<mpfr_float>>(b_);
Vec<mpfr_float>& bstar = std::get<Vec<mpfr_float>>(b_minus_bstar_);
Vec<mpfr_float>& c = std::get<Vec<mpfr_float>>(c_);
assert(Precision(dhdttemp) == current_precision_);
assert(Precision(dhdx0) == current_precision_);
assert(Precision(dhdxtemp) == current_precision_);
assert(Precision(a) == current_precision_);
assert(Precision(b) == current_precision_);
if (uses_embedded_) assert(Precision(bstar) == current_precision_);
assert(Precision(c) == current_precision_);
}
/**
\brief Perform a generic predictor step.
\param next_space The computed prediction.
\param method An enum class selecting the predictor method to use.
\param S The system being solved.
\param current_space The current space variable vector.
\param current_time The current time.
\param delta_t The size of the time step.
\param condition_number_estimate The computed estimate of the condition
number of the Jacobian. \param
num_steps_since_last_condition_number_computation. Updated in this function.
\param frequency_of_CN_estimation How many steps to take between condition
number estimates. \param prec_type The operating precision type. \param
tracking_tolerance How tightly to track the path.
\return SuccessCode indicating how the prediction went.
*/
template <typename ComplexType>
SuccessCode Predict(
Vec<ComplexType>& next_space, System const& S,
const Vec<ComplexType>& current_space, ComplexType current_time,
ComplexType const& delta_t, NumErrorT& condition_number_estimate,
unsigned& num_steps_since_last_condition_number_computation,
unsigned frequency_of_CN_estimation,
NumErrorT const& tracking_tolerance) {
auto step_success =
FullStep(next_space, S, current_space, current_time, delta_t);
NumErrorT norm_J, norm_J_inverse;
SetNormsCond<ComplexType>(norm_J, norm_J_inverse, condition_number_estimate,
num_steps_since_last_condition_number_computation,
frequency_of_CN_estimation);
return step_success;
}
/**
\brief Perform a generic predictor step and return size_proportion and
condition number information
\param next_space The computed prediction.
\param method An enum class selecting the predictor method to use.
\param size_proportion $a$ in AMP2 paper.
\param norm_J The computed estimate of the norm of the Jacobian matrix.
\param norm_J_inverse The computed estimate of the norm of the inverse of the
Jacobian matrix. \param S The system being solved. \param current_space The
current space variable vector. \param current_time The current time. \param
delta_t The size of the time step. \param condition_number_estimate The
computed estimate of the condition number of the Jacobian. \param
num_steps_since_last_condition_number_computation. Updated in this function.
\param frequency_of_CN_estimation How many steps to take between condition
number estimates. \param prec_type The operating precision type. \param
tracking_tolerance How tightly to track the path. \param AMP_config The
settings for adaptive multiple precision.
\return SuccessCode indicating how the prediction went.
*/
template <typename ComplexType>
SuccessCode Predict(
Vec<ComplexType>& next_space, NumErrorT& size_proportion,
NumErrorT& norm_J, NumErrorT& norm_J_inverse, System const& S,
const Vec<ComplexType>& current_space, ComplexType current_time,
ComplexType const& delta_t, NumErrorT& condition_number_estimate,
unsigned& num_steps_since_last_condition_number_computation,
unsigned frequency_of_CN_estimation, NumErrorT const& tracking_tolerance,
AdaptiveMultiplePrecisionConfig const& AMP_config) {
auto success_code =
Predict<ComplexType>(next_space, S, current_space, current_time,
delta_t, condition_number_estimate,
num_steps_since_last_condition_number_computation,
frequency_of_CN_estimation, tracking_tolerance);
SetNormsCond<ComplexType>(norm_J, norm_J_inverse, condition_number_estimate,
num_steps_since_last_condition_number_computation,
frequency_of_CN_estimation);
// Set size_proportion
SetSizeProportion(size_proportion, delta_t);
if (success_code != SuccessCode::Success) return success_code;
// AMP Criteria
if (!amp::CriterionA<ComplexType>(norm_J, norm_J_inverse,
AMP_config)) // AMP_criterion_A != ok
{
return SuccessCode::HigherPrecisionNecessary;
} else if (!amp::CriterionC<ComplexType>(
norm_J_inverse, current_space, tracking_tolerance,
AMP_config)) // AMP_criterion_C != ok
{
return SuccessCode::HigherPrecisionNecessary;
}
return success_code;
}
/**
\brief Perform a generic predictor step and return error estimate,
size_proportion and condition number information
\param next_space The computed prediction.
\param method An enum class selecting the predictor method to use.
\param error_estimate Estimate of the error from an embedded method.
\param size_proportion $a$ in AMP2 paper.
\param norm_J The computed estimate of the norm of the Jacobian matrix.
\param norm_J_inverse The computed estimate of the norm of the inverse of the
Jacobian matrix. \param S The system being solved. \param current_space The
current space variable vector. \param current_time The current time. \param
delta_t The size of the time step. \param condition_number_estimate The
computed estimate of the condition number of the Jacobian. \param
num_steps_since_last_condition_number_computation. Updated in this function.
\param frequency_of_CN_estimation How many steps to take between condition
number estimates. \param prec_type The operating precision type. \param
tracking_tolerance How tightly to track the path. \param AMP_config The
settings for adaptive multiple precision.
\return SuccessCode indicating how the prediction went.
*/
template <typename ComplexType>
SuccessCode Predict(
Vec<ComplexType>& next_space, NumErrorT& error_estimate,
NumErrorT& size_proportion, NumErrorT& norm_J, NumErrorT& norm_J_inverse,
System const& S, const Vec<ComplexType>& current_space,
ComplexType current_time, ComplexType const& delta_t,
NumErrorT& condition_number_estimate,
unsigned& num_steps_since_last_condition_number_computation,
unsigned frequency_of_CN_estimation, NumErrorT const& tracking_tolerance,
AdaptiveMultiplePrecisionConfig const& AMP_config) {
// If this is a method without an error estimator, then can't calculate size
// proportion and should throw an error
if (!predict::HasErrorEstimate(predictor_))
throw std::runtime_error(
"incompatible predictor choice in ExplicitPredict, no error "
"estimator");
auto success_code =
Predict(next_space, size_proportion, norm_J, norm_J_inverse, S,
current_space, current_time, delta_t, condition_number_estimate,
num_steps_since_last_condition_number_computation,
frequency_of_CN_estimation, tracking_tolerance, AMP_config);
SetErrorEstimate(error_estimate, delta_t);
return success_code;
}
/**
\brief Get the currently used prediction method.
\return The current method.
*/
Predictor PredictorMethod() { return predictor_; }
/**
\brief Get the order of the currently used prediction method.
This is the lowest order of the predictor. The order of the error estimate is
this plus one.
\return The aforementioned order.
*/
inline unsigned Order() { return p_; }
/**
\brief Get whether the current prediction method provides an error estimate.
\return Yes or no.
*/
inline bool HasErrorEstimate() {
return predict::HasErrorEstimate(predictor_);
}
protected:
///////////////////////////
//
// Protected Methods
//
////////////////////
template <typename T>
Eigen::PartialPivLU<Mat<T>>& GetLU() {
return LUSelector<T>::Run(*this);
}
Eigen::PartialPivLU<Mat<dbl>>& GetLU_d() { return LU_d_; }
Eigen::PartialPivLU<Mat<mpfr_complex>>& GetLU_mp() {
assert(current_precision_ == DefaultPrecision());
return LU_mp_[current_precision_];
}
/**
\brief Performs a full prediction step from current_time to current_time +
delta_t
\param next_space The computed prediction space
\param S The homotopy system
\param current_space The current space values
\param current_time The current time values
\param delta_t The time step
\return SuccessCode determining result of the computation
*/
template <typename ComplexType>
SuccessCode FullStep(Vec<ComplexType>& next_space, System const& S,
Vec<ComplexType> const& current_space,
ComplexType const& current_time,
ComplexType const& delta_t) {
// If using constant predictor
if (s_ == 0) {
next_space = current_space;
return SuccessCode::Success;
}
using RealType = typename Eigen::NumTraits<ComplexType>::Real;
Mat<ComplexType>& Kref = std::get<Mat<ComplexType>>(K_);
Mat<RealType>& aref = std::get<Mat<RealType>>(a_);
Vec<RealType>& bref = std::get<Vec<RealType>>(b_);
Vec<RealType>& cref = std::get<Vec<RealType>>(c_);
Kref.fill(ComplexType(0));
Vec<ComplexType> temp(S.NumTotalFunctions());
if (EvalRHS(S, current_space, current_time, Kref, 0) !=
SuccessCode::Success) {
return SuccessCode::MatrixSolveFailureFirstPartOfPrediction;
}
for (int ii = 1; ii < s_; ++ii) {
temp.setZero(); // see https://github.com/bertiniteam/b2/issues/198
for (int jj = 0; jj < ii; ++jj) temp += aref(ii, jj) * Kref.col(jj);
// Vec<ComplexType> wfp =
if (EvalRHS<ComplexType>(S, current_space + delta_t * temp,
current_time + cref(ii) * delta_t, Kref,
ii) != SuccessCode::Success)
return SuccessCode::MatrixSolveFailure;
}
temp.setZero();
for (int ii = 0; ii < s_; ++ii) temp += bref(ii) * Kref.col(ii);
next_space = current_space + delta_t * temp;
return SuccessCode::Success;
};
template <typename ComplexType>
void SetNormsCond(NumErrorT& norm_J, NumErrorT& norm_J_inverse,
NumErrorT& condition_number_estimate,
unsigned num_steps_since_last_condition_number_computation,
unsigned frequency_of_CN_estimation) {
// Calculate condition number and update if needed
Eigen::PartialPivLU<Mat<ComplexType>>& LUref = GetLU<ComplexType>();
Mat<ComplexType>& dhdxref = std::get<Mat<ComplexType>>(dh_dx_0_);
// TODO this random vector should not be made fresh every time. especiallyif
// the numeric type is mpfr_complex!
Vec<ComplexType> randy = RandomOfUnits<ComplexType>(numVariables_);
Vec<ComplexType> temp_soln = LUref.solve(randy);
norm_J = NumErrorT(dhdxref.norm());
norm_J_inverse = NumErrorT(temp_soln.norm());
if (num_steps_since_last_condition_number_computation >=
frequency_of_CN_estimation) {
condition_number_estimate = NumErrorT(norm_J * norm_J_inverse);
num_steps_since_last_condition_number_computation =
1; // reset the counter to 1
} else // no need to compute the condition number
num_steps_since_last_condition_number_computation++;
}
/**
\brief Computes the error estimate of this prediction step.
\param error_estimate Computed error estimate
\param delta_t The time step
\return Success code or the computation
*/
template <typename ComplexType>
SuccessCode SetErrorEstimate(NumErrorT& error_estimate,
ComplexType const& delta_t) {
using RealType = typename Eigen::NumTraits<ComplexType>::Real;
Mat<ComplexType>& Kref = std::get<Mat<ComplexType>>(K_);
Vec<RealType>& b_minus_bstar_ref = std::get<Vec<RealType>>(b_minus_bstar_);
auto numFuncs = Kref.rows();
Vec<ComplexType> err(numFuncs);
err.setZero();
for (int ii = 0; ii < s_; ++ii) {
err += (b_minus_bstar_ref(ii)) * Kref.col(ii);
}
err *= delta_t;
error_estimate = NumErrorT(err.norm());
return SuccessCode::Success;
};
/**
\brief Compute the size proportion variable for AMP computation
\param size_proportion Computed size proportion
\param delta_t The time step
\return Success code of the computation
*/
template <typename ComplexType>
SuccessCode SetSizeProportion(NumErrorT& size_proportion,
ComplexType const& delta_t) {
if (predict::HasErrorEstimate(predictor_)) {
NumErrorT err_est;
SetErrorEstimate(err_est, delta_t);
using std::pow;
size_proportion = err_est / NumErrorT(pow(abs(delta_t), p_ + 1));
return SuccessCode::Success;
} else {
Mat<ComplexType>& Kref = std::get<Mat<ComplexType>>(K_);
using std::pow;
size_proportion =
NumErrorT(Kref.array().abs().maxCoeff() / (pow(abs(delta_t), p_)));
return SuccessCode::Success;
}
};
/**
\brief Evaluates the RHS of the Davidenko differential equation at a
particular time and space
\param S The homotopy system
\param space The space variable used to evaluate RHS
\param time The time variable used to evaluate RHS
\param K Matrix of stage variables
\param stage Which stage variable(column of K) should be filled by this
computation
\return Success code of this computation
*/
template <typename ComplexType>
SuccessCode EvalRHS(System const& S, const Vec<ComplexType>& space,
const ComplexType& time, Mat<ComplexType>& K,
unsigned stage) {
if (std::is_same<ComplexType, mpfr_complex>::value) PrecisionSanityCheck();
if (stage == 0) {
Eigen::PartialPivLU<Mat<ComplexType>>& LUref = GetLU<ComplexType>();
Mat<ComplexType>& dhdxref = std::get<Mat<ComplexType>>(dh_dx_0_);
if (!std::is_same<ComplexType, dbl>::value) {
assert(DefaultPrecision() == current_precision_);
assert(Precision(space) == current_precision_);
assert(Precision(time) == current_precision_);
assert(Precision(dhdxref) == current_precision_);
assert(Precision(K) == current_precision_);
}
S.SetAndReset<ComplexType>(space, time);
S.JacobianInPlace(dhdxref);
LUref = dhdxref.lu();
if (!std::is_same<ComplexType, dbl>::value) {
assert(Precision(dhdxref) == current_precision_);
assert(Precision(LUref.matrixLU()) == current_precision_);
}
if (LUPartialPivotDecompositionSuccessful(LUref.matrixLU()) !=
MatrixSuccessCode::Success)
return SuccessCode::MatrixSolveFailureFirstPartOfPrediction;
Vec<ComplexType>& dhdtref = std::get<Vec<ComplexType>>(dh_dt_temp_);
S.TimeDerivativeInPlace(dhdtref);
K.col(stage) = LUref.solve(-dhdtref);
return SuccessCode::Success;
} else {
S.SetAndReset<ComplexType>(space, time);
Mat<ComplexType>& dhdxtempref = std::get<Mat<ComplexType>>(dh_dx_temp_);
S.JacobianInPlace(dhdxtempref);
auto LU = dhdxtempref.lu();
if (LUPartialPivotDecompositionSuccessful(LU.matrixLU()) !=
MatrixSuccessCode::Success)
return SuccessCode::MatrixSolveFailure;
Vec<ComplexType>& dhdtref = std::get<Vec<ComplexType>>(dh_dt_temp_);
S.TimeDerivativeInPlace(dhdtref);
K.col(stage) = LU.solve(-dhdtref);
return SuccessCode::Success;
}
}
private:
///////////////////////////
//
// Private Methods
//
////////////////////
/**
/brief Fills the local embedded butcher table variables a,b,bstar and c with
the constant static values stored in the class.
\param stages Number of stages. Used to create correct size on variables
\param a Matrix of the Butcher table
\param b Weights in the Butcher table
\param bstar Weights of embedded method in Butcher table
\param c Time offsets in Butcher table
*/
template <typename RealType>
void FillButcherTable(int stages, const Mat<mpq_rational>& a,
const Mat<mpq_rational>& b,
const Mat<mpq_rational>& b_minus_bstar,
const Mat<mpq_rational>& c) {
Mat<RealType>& aref = std::get<Mat<RealType>>(a_);
aref.resize(stages, stages);
for (int ii = 0; ii < stages; ++ii) {
for (int jj = 0; jj < s_; ++jj) {
aref(ii, jj) = static_cast<RealType>(a(ii, jj));
}
}
Vec<RealType>& bref = std::get<Vec<RealType>>(b_);
bref.resize(stages);
for (int ii = 0; ii < stages; ++ii) {
bref(ii) = static_cast<RealType>(b(ii));
}
Vec<RealType>& b_minus_bstar_ref = std::get<Vec<RealType>>(b_minus_bstar_);
b_minus_bstar_ref.resize(stages);
for (int ii = 0; ii < stages; ++ii) {
b_minus_bstar_ref(ii) = static_cast<RealType>(b_minus_bstar(ii));
}
Vec<RealType>& cref = std::get<Vec<RealType>>(c_);
cref.resize(stages);
for (int ii = 0; ii < stages; ++ii) {
cref(ii) = static_cast<RealType>(c(ii));
}
uses_embedded_ = true;
}
/**
/brief Fills the local butcher table variables a,b and c with the constant
static values stored in the class.
\param stages Number of stages. Used to create correct size on variables
\param a Matrix of the Butcher table
\param b Weights in the Butcher table
\param c Time offsets in Butcher table
*/
template <typename RealType>
void FillButcherTable(int stages, const Mat<mpq_rational>& a,
const Mat<mpq_rational>& b,
const Mat<mpq_rational>& c) {
Mat<RealType>& aref = std::get<Mat<RealType>>(a_);
aref.resize(stages, stages);
for (int ii = 0; ii < stages; ++ii) {
for (int jj = 0; jj < s_; ++jj) {
aref(ii, jj) = static_cast<RealType>(a(ii, jj));
}
}
Vec<RealType>& bref = std::get<Vec<RealType>>(b_);
bref.resize(stages);
for (int ii = 0; ii < stages; ++ii) {
bref(ii) = static_cast<RealType>(b(ii));
}
Vec<RealType>& cref = std::get<Vec<RealType>>(c_);
cref.resize(stages);
for (int ii = 0; ii < stages; ++ii) {
cref(ii) = static_cast<RealType>(c(ii));
}
uses_embedded_ = false;
}
///////////////////////////
//
// Private Data Members
//
////////////////////
unsigned
numTotalFunctions_; // Number of total functions for the current system
unsigned numVariables_; // Number of variables for the current system
mutable std::tuple<Mat<dbl>, Mat<mpfr_complex>>
K_; // All the stage variables. Each column represents a different
// stage.
Predictor predictor_; // Method for prediction
unsigned p_; // Order of the prediction method
mutable std::tuple<Mat<dbl>, Mat<mpfr_complex>>
dh_dx_0_; // Jacobian for the initial stage. Use for AMP testing
mutable std::tuple<Mat<dbl>, Mat<mpfr_complex>>
dh_dx_temp_; // Temporary jacobian for all other stages
mutable std::tuple<Vec<dbl>, Vec<mpfr_complex>>
dh_dt_temp_; // Temporary time derivative used for all stages
// std::tuple< Eigen::PartialPivLU<Mat<dbl>>,
// Eigen::PartialPivLU<Mat<mpfr_complex>> > LU_0_; // LU from the intial
// stage used for AMP testing
mutable Eigen::PartialPivLU<Mat<dbl>> LU_d_;
mutable std::map<unsigned, Eigen::PartialPivLU<Mat<mpfr_complex>>> LU_mp_;
// Butcher Table (notation from
// https://en.wikipedia.org/wiki/List_of_Runge%E2%80%93Kutta_methods )
mutable unsigned s_; // Number of stages
mutable std::tuple<Mat<double>, Mat<mpfr_float>> a_;
mutable std::tuple<Vec<double>, Vec<mpfr_float>> b_;
mutable std::tuple<Vec<double>, Vec<mpfr_float>> b_minus_bstar_;
mutable std::tuple<Vec<double>, Vec<mpfr_float>> c_;
mutable bool uses_embedded_;
mutable unsigned current_precision_;
// static const variables that store the Butcher table in mpq_rational form
// Euler
static const mpq_rational aEulerPtr_[];
static const Eigen::Matrix<mpq_rational, 1, 1> aEuler_;
static const mpq_rational bEulerPtr_[];
static const Eigen::Matrix<mpq_rational, 1, 1> bEuler_;
static const mpq_rational cEulerPtr_[];
static const Eigen::Matrix<mpq_rational, 1, 1> cEuler_;
// Heun-Euler
static const mpq_rational aHeunEulerPtr_[];
static const Eigen::Matrix<mpq_rational, 2, 2> aHeunEuler_;
static const mpq_rational bHeunEulerPtr_[];
static const Eigen::Matrix<mpq_rational, 2, 1> bHeunEuler_;
static const mpq_rational b_minus_bstarHeunEulerPtr_[];
static const Eigen::Matrix<mpq_rational, 2, 1> b_minus_bstarHeunEuler_;
static const mpq_rational cHeunEulerPtr_[];
static const Eigen::Matrix<mpq_rational, 2, 1> cHeunEuler_;
// RK4
static const mpq_rational aRK4Ptr_[];
static const Eigen::Matrix<mpq_rational, 4, 4> aRK4_;
static const mpq_rational bRK4Ptr_[];
static const Eigen::Matrix<mpq_rational, 4, 1> bRK4_;
static const mpq_rational cRK4Ptr_[];
static const Eigen::Matrix<mpq_rational, 4, 1> cRK4_;
// RKF45
static const mpq_rational aRKF45Ptr_[];
static const Eigen::Matrix<mpq_rational, 6, 6> aRKF45_;
static const mpq_rational bRKF45Ptr_[];
static const Eigen::Matrix<mpq_rational, 6, 1> bRKF45_;
static const mpq_rational b_minus_bstarRKF45Ptr_[];
static const Eigen::Matrix<mpq_rational, 6, 1> b_minus_bstarRKF45_;
static const mpq_rational cRKF45Ptr_[];
static const Eigen::Matrix<mpq_rational, 6, 1> cRKF45_;
// RK Cash-Karp45
static const mpq_rational aRKCK45Ptr_[];
static const Eigen::Matrix<mpq_rational, 6, 6> aRKCK45_;
static const mpq_rational bRKCK45Ptr_[];
static const Eigen::Matrix<mpq_rational, 6, 1> bRKCK45_;
static const mpq_rational b_minus_bstarRKCK45Ptr_[];
static const Eigen::Matrix<mpq_rational, 6, 1> b_minus_bstarRKCK45_;
static const mpq_rational cRKCK45Ptr_[];
static const Eigen::Matrix<mpq_rational, 6, 1> cRKCK45_;
// RK Dormand-Prince 56
static const mpq_rational aRKDP56Ptr_[];
static const Eigen::Matrix<mpq_rational, 8, 8> aRKDP56_;
static const mpq_rational bRKDP56Ptr_[];
static const Eigen::Matrix<mpq_rational, 8, 1> bRKDP56_;
static const mpq_rational b_minus_bstarRKDP56Ptr_[];
static const Eigen::Matrix<mpq_rational, 8, 1> b_minus_bstarRKDP56_;
static const mpq_rational cRKDP56Ptr_[];
static const Eigen::Matrix<mpq_rational, 8, 1> cRKDP56_;
// RK Verner 67
static const mpq_rational aRKV67Ptr_[];
static const Eigen::Matrix<mpq_rational, 10, 10> aRKV67_;
static const mpq_rational bRKV67Ptr_[];
static const Eigen::Matrix<mpq_rational, 10, 1> bRKV67_;
static const mpq_rational b_minus_bstarRKV67Ptr_[];
static const Eigen::Matrix<mpq_rational, 10, 1> b_minus_bstarRKV67_;
static const mpq_rational cRKV67Ptr_[];
static const Eigen::Matrix<mpq_rational, 10, 1> cRKV67_;
}; // re: ExplicitRKPredictor class
} // namespace predict
} // namespace tracking
} // namespace bertini
#endif