ObjectiveFunctionTests.h

Go to the documentation of this file.

/*
    Copyright (C) 2011, Hammersmith Imanet Ltd
    Copyright (C) 2013, 2020, 2022-2024 University College London
    This file is part of STIR.
 
    SPDX-License-Identifier: Apache-2.0
 
    See STIR/LICENSE.txt for details
*/
 
#include "stir/RunTests.h"
#include "stir/IO/write_to_file.h"
#include "stir/info.h"
#include "stir/Succeeded.h"
#include <iostream>
 
START_NAMESPACE_STIR

template <class ObjectiveFunctionT, class TargetT>
class ObjectiveFunctionTests : public RunTests
{
public:
  typedef ObjectiveFunctionT objective_function_type;
  typedef TargetT target_type;


  virtual Succeeded test_gradient(const std::string& test_name,
                                  ObjectiveFunctionT& objective_function,
                                  TargetT& target,
                                  const float eps,
                                  const bool full_gradient = true);


  virtual Succeeded
  test_Hessian(const std::string& test_name, ObjectiveFunctionT& objective_function, const TargetT& target, const float eps);
  virtual Succeeded test_Hessian_concavity(const std::string& test_name,
                                           ObjectiveFunctionT& objective_function,
                                           const TargetT& target,
                                           const float mult_factor = 1.F);
 
protected:

  virtual shared_ptr<const TargetT> construct_increment(const TargetT& target, const float eps) const;
};
 
template <class ObjectiveFunctionT, class TargetT>
Succeeded
ObjectiveFunctionTests<ObjectiveFunctionT, TargetT>::test_gradient(const std::string& test_name,
                                                                   ObjectiveFunctionT& objective_function,
                                                                   TargetT& target,
                                                                   const float eps,
                                                                   const bool full_gradient)
{
  shared_ptr<TargetT> gradient_sptr(target.get_empty_copy());
  shared_ptr<TargetT> gradient_2_sptr(target.get_empty_copy());
  info("Computing gradient");
  objective_function.compute_gradient(*gradient_sptr, target);
  this->set_tolerance(std::max(fabs(double(gradient_sptr->find_min())), double(gradient_sptr->find_max())) / 1000);
  info("Computing objective function at target");
  const double value_at_target = objective_function.compute_value(target);
  bool testOK = true;
  if (full_gradient)
    {
      info("Computing gradient of objective function by numerical differences (this will take a while)");
      auto target_iter = target.begin_all();
      auto gradient_iter = gradient_sptr->begin_all();
      auto gradient_2_iter = gradient_2_sptr->begin_all();
      while (target_iter != target.end_all())
        {
          *target_iter += eps;
          const double value_at_inc = objective_function.compute_value(target);
          *target_iter -= eps;
          const float gradient_at_iter = static_cast<float>((value_at_inc - value_at_target) / eps);
          *gradient_2_iter++ = gradient_at_iter;
          testOK = testOK && this->check_if_equal(gradient_at_iter, *gradient_iter, "gradient");
          ++target_iter;
          ++gradient_iter;
        }
    }
  else
    {
      /* test f(x+dx) - f(x) ~ dx^t G(x) */
      shared_ptr<const TargetT> increment_sptr = this->construct_increment(target, eps);
      shared_ptr<TargetT> target_plus_inc_sptr(target.clone());
      *target_plus_inc_sptr += *increment_sptr;
      const double value_at_inc = objective_function.compute_value(*target_plus_inc_sptr);
      const double my_sum = std::inner_product(
          gradient_sptr->begin_all_const(), gradient_sptr->end_all_const(), increment_sptr->begin_all_const(), 0.);
 
      testOK = testOK && this->check_if_equal(value_at_inc - value_at_target, my_sum, "gradient");
    }
 
  if (!testOK)
    {
      std::cerr << "Numerical gradient test failed with for " + test_name + "\n";
      std::cerr << "Writing diagnostic files " << test_name
                << "_target.hv, *gradient.hv (and *numerical_gradient.hv if full gradient test is used)\n";
      write_to_file(test_name + "_target.hv", target);
      write_to_file(test_name + "_gradient.hv", *gradient_sptr);
      if (full_gradient)
        write_to_file(test_name + "_numerical_gradient.hv", *gradient_2_sptr);
      return Succeeded::no;
    }
  else
    {
      return Succeeded::yes;
    }
}
 
template <class ObjectiveFunctionT, class TargetT>
shared_ptr<const TargetT>
ObjectiveFunctionTests<ObjectiveFunctionT, TargetT>::construct_increment(const TargetT& target, const float eps) const
{
  shared_ptr<TargetT> increment_sptr(target.clone());
  *increment_sptr *= eps / increment_sptr->find_max();
  *increment_sptr += eps / 2;
  return increment_sptr;
}
 
template <class ObjectiveFunctionT, class TargetT>
Succeeded
ObjectiveFunctionTests<ObjectiveFunctionT, TargetT>::test_Hessian(const std::string& test_name,
                                                                  ObjectiveFunctionT& objective_function,
                                                                  const TargetT& target,
                                                                  const float eps)
{
  info("Comparing Hessian*dx with difference of gradients");
 
  /* test G(x+dx) = G(x) + H dx + small stuff */
  shared_ptr<TargetT> gradient_sptr(target.get_empty_copy());
  shared_ptr<TargetT> gradient_2_sptr(target.get_empty_copy());
  shared_ptr<TargetT> output(target.get_empty_copy());
  shared_ptr<const TargetT> increment_sptr = this->construct_increment(target, eps);
  shared_ptr<TargetT> target_plus_inc_sptr(target.clone());
  *target_plus_inc_sptr += *increment_sptr;
 
  info("Computing gradient");
  objective_function.compute_gradient(*gradient_sptr, target);
  objective_function.compute_gradient(*gradient_2_sptr, *target_plus_inc_sptr);
  this->set_tolerance(std::max(fabs(double(gradient_sptr->find_min())), double(gradient_sptr->find_max())) / 1E5);
  info("Computing Hessian * increment at target");
  objective_function.accumulate_Hessian_times_input(*output, target, *increment_sptr);
  auto output_iter = output->begin_all_const();
  auto gradient_iter = gradient_sptr->begin_all_const();
  auto gradient_2_iter = gradient_2_sptr->begin_all_const();
  bool testOK = true;
  while (output_iter != output->end_all())
    {
      testOK = testOK && this->check_if_equal(*gradient_2_iter - *gradient_iter, *output_iter, "Hessian*increment");
      ++output_iter;
      ++gradient_iter;
      ++gradient_2_iter;
    }
  if (!testOK)
    {
      std::cerr << "Numerical Hessian test failed with for " + test_name + "\n";
      std::cerr << "Writing diagnostic files " << test_name
                << "_target.hv, *gradient.hv, *increment, *numerical_gradient.hv, *Hessian_times_increment\n";
      write_to_file(test_name + "_target.hv", target);
      write_to_file(test_name + "_gradient.hv", *gradient_sptr);
      write_to_file(test_name + "_increment.hv", *increment_sptr);
      write_to_file(test_name + "_gradient_at_increment.hv", *gradient_2_sptr);
      write_to_file(test_name + "_Hessian_times_increment.hv", *output);
      return Succeeded::no;
    }
  else
    {
      return Succeeded::yes;
    }
}
 
template <class ObjectiveFunctionT, class TargetT>
Succeeded
ObjectiveFunctionTests<ObjectiveFunctionT, TargetT>::test_Hessian_concavity(const std::string& test_name,
                                                                            ObjectiveFunctionT& objective_function,
                                                                            const TargetT& target,
                                                                            const float mult_factor)
{
  shared_ptr<TargetT> output(target.get_empty_copy());

  objective_function.accumulate_Hessian_times_input(*output, target, target);

  const float my_sum = std::inner_product(target.begin_all(), target.end_all(), output->begin_all(), 0.F) * mult_factor;
 
  // test for a CONCAVE function (after multiplying with mult_factor of course)
  if (this->check_if_less(my_sum, 0))
    {
      return Succeeded::yes;
    }
  else
    {
      // print to console the FAILED configuration
      std::cerr << "FAIL: " + test_name + ": Computation of x^T H x = " + std::to_string(my_sum)
                << " > 0 (Hessian) and is therefore NOT concave"
                << "\n >target image max=" << target.find_max() << "\n >target image min=" << target.find_min()
                << "\n >output image max=" << output->find_max() << "\n >output image min=" << output->find_min() << '\n';
      std::cerr << "Writing diagnostic files to " << test_name + "_concavity_out.hv, *target.hv\n";
      write_to_file(test_name + "_concavity_out.hv", *output);
      write_to_file(test_name + "_target.hv", target);
      return Succeeded::no;
    }
}
 
END_NAMESPACE_STIR