diff --git a/src/LearningMethods/GradientDescent.cpp b/src/LearningMethods/GradientDescent.cpp
index 5727490f88d5dba8d59a8ea929f3de2fcdf2ae25..3832a2328fa836a6015fa8e5533da1a88b58af7b 100644
--- a/src/LearningMethods/GradientDescent.cpp
+++ b/src/LearningMethods/GradientDescent.cpp
@@ -13,16 +13,16 @@ namespace lib4neuro {
GradientDescent::GradientDescent(double epsilon, size_t n_to_restart, int max_iters, size_t batch) {
this->tolerance = epsilon;
this->restart_frequency = n_to_restart;
- this->optimal_parameters = new std::vector<double>(0);
+// this->optimal_parameters = new std::vector<double>(0);
this->maximum_niters = max_iters;
this->batch = batch;
}
GradientDescent::~GradientDescent() {
- if (this->optimal_parameters) {
- delete this->optimal_parameters;
- this->optimal_parameters = nullptr;
- }
+// if (this->optimal_parameters) {
+// delete this->optimal_parameters;
+// this->optimal_parameters = nullptr;
+// }
}
void GradientDescent::eval_step_size_mk(double &gamma,
@@ -295,8 +295,8 @@ namespace lib4neuro {
#endif
}
- *this->optimal_parameters = *params_current;
- ef.get_network_instance()->copy_parameter_space( this->optimal_parameters );
+ this->optimal_parameters = *params_current;
+ ef.get_network_instance()->copy_parameter_space( &this->optimal_parameters );
delete gradient_current;
delete gradient_prev;
@@ -305,6 +305,6 @@ namespace lib4neuro {
}
std::vector<double> *GradientDescent::get_parameters() {
- return this->optimal_parameters;
+ return &this->optimal_parameters;
}
}
diff --git a/src/LearningMethods/GradientDescent.h b/src/LearningMethods/GradientDescent.h
index 352def273f96e59c5098d0e977d0a6548493eaa6..325078b00dc6017d9919258e8e2f677040c4f293 100644
--- a/src/LearningMethods/GradientDescent.h
+++ b/src/LearningMethods/GradientDescent.h
@@ -44,7 +44,7 @@ namespace lib4neuro {
/**
* Vector of minima coordinates
*/
- std::vector<double> *optimal_parameters;
+ std::vector<double> optimal_parameters;
/**
* Adaptive calculation of the step-size based on several historical characteristics.
diff --git a/src/LearningMethods/LearningSequence.cpp b/src/LearningMethods/LearningSequence.cpp
index 8d5f9ce7f88e6fe31ff6fb69e6b8a03fb1248b25..05569b85de4b68e797a1e4db4928f777c017f68f 100644
--- a/src/LearningMethods/LearningSequence.cpp
+++ b/src/LearningMethods/LearningSequence.cpp
@@ -13,13 +13,10 @@ namespace lib4neuro {
LearningSequence::LearningSequence( double tolerance, int max_n_cycles ){
this->tol = tolerance;
this->max_number_of_cycles = max_n_cycles;
- this->best_parameters = new std::vector<double>();
+// this->best_parameters = new std::vector<double>();
}
- LearningSequence::~LearningSequence() {
-
-
- }
+ LearningSequence::~LearningSequence() = default;
std::vector<double>* LearningSequence::get_parameters() {
if( this->learning_sequence.size() > 0 ){
@@ -28,7 +25,7 @@ namespace lib4neuro {
return nullptr;
}
- void LearningSequence::add_learning_method(LearningMethod *method) {
+ void LearningSequence::add_learning_method(std::shared_ptr<LearningMethod> method) {
this->learning_sequence.push_back( method );
}
@@ -49,16 +46,16 @@ namespace lib4neuro {
if( error < the_best_error ){
the_best_error = error;
- *this->best_parameters = *ef.get_parameters();
+ this->best_parameters = *ef.get_parameters();
}
if( error <= this->tol ){
- ef.get_network_instance()->copy_parameter_space( this->best_parameters );
+ ef.get_network_instance()->copy_parameter_space( &this->best_parameters );
return;
}
}
COUT_DEBUG("Cycle: " << cycle_idx << ", the lowest error: " << the_best_error << std::endl );
}
- ef.get_network_instance()->copy_parameter_space( this->best_parameters );
+ ef.get_network_instance()->copy_parameter_space( &this->best_parameters );
}
}
\ No newline at end of file
diff --git a/src/LearningMethods/LearningSequence.h b/src/LearningMethods/LearningSequence.h
index bb97591093d0504099fb1bf078b4ab3629679297..3f3f4b5d0c01c0861c365c05451006ddaf227395 100644
--- a/src/LearningMethods/LearningSequence.h
+++ b/src/LearningMethods/LearningSequence.h
@@ -24,7 +24,7 @@ namespace lib4neuro {
/**
*
*/
- std::vector<LearningMethod*> learning_sequence;
+ std::vector<std::shared_ptr<LearningMethod>> learning_sequence;
/**
*
@@ -34,7 +34,7 @@ namespace lib4neuro {
/**
*
*/
- std::vector<double> *best_parameters = nullptr;
+ std::vector<double> best_parameters; // = nullptr;
/**
*
@@ -71,7 +71,7 @@ namespace lib4neuro {
*
* @param method
*/
- LIB4NEURO_API void add_learning_method( LearningMethod * method );
+ LIB4NEURO_API void add_learning_method( std::shared_ptr<LearningMethod> method );
};
}
diff --git a/src/LearningMethods/LevenbergMarquardt.cpp b/src/LearningMethods/LevenbergMarquardt.cpp
index b9db3539c49bc56087f010fc34fd908b41b91cc2..d10eab7b2207ad46844378b67729edac82e11df8 100644
--- a/src/LearningMethods/LevenbergMarquardt.cpp
+++ b/src/LearningMethods/LevenbergMarquardt.cpp
@@ -31,7 +31,7 @@ struct lib4neuro::LevenbergMarquardt::LevenbergMarquardtImpl {
/**
* Vector of minimum coordinates
*/
- std::vector<double> *optimal_parameters;
+ std::vector<double> optimal_parameters;
/**
* Returns Jacobian matrix of the residual function using the backpropagation algorithm
@@ -92,7 +92,7 @@ namespace lib4neuro {
this->p_impl->lambda_increase = lambda_increase;
this->p_impl->lambda_decrease = lambda_decrease;
this->p_impl->maximum_niters = max_iters;
- this->p_impl->optimal_parameters = new std::vector<double>();
+ this->p_impl->optimal_parameters; // = new std::vector<double>();
}
void LevenbergMarquardt::optimize(lib4neuro::ErrorFunction& ef,
@@ -229,16 +229,16 @@ namespace lib4neuro {
COUT_DEBUG("Iteration: " << iter_counter << " Current error: " << current_err << ", Current gradient norm: " << gradient_norm << ", Direction norm: " << update_norm << std::endl);
/* Store the optimized parameters */
- *this->p_impl->optimal_parameters = *params_current;
+ this->p_impl->optimal_parameters = *params_current;
- ef.get_network_instance()->copy_parameter_space(this->p_impl->optimal_parameters);
+ ef.get_network_instance()->copy_parameter_space(&this->p_impl->optimal_parameters);
delete params_tmp;
}
std::vector<double>* LevenbergMarquardt::get_parameters() {
- return this->p_impl->optimal_parameters;
+ return &this->p_impl->optimal_parameters;
}
LevenbergMarquardt::~LevenbergMarquardt() = default;
diff --git a/src/LearningMethods/RandomSolution.cpp b/src/LearningMethods/RandomSolution.cpp
index 84cabc9e347cf39da3bac71a24e6ec1a8c87ef78..f5362566b1ff8c99f2ccd10628e174dba7caea01 100644
--- a/src/LearningMethods/RandomSolution.cpp
+++ b/src/LearningMethods/RandomSolution.cpp
@@ -11,21 +11,21 @@
namespace lib4neuro {
RandomSolution::RandomSolution() {
- this->optimal_parameters = new std::vector<double>();
+// this->optimal_parameters = new std::vector<double>();
}
RandomSolution::~RandomSolution() {
}
- std::vector<double> *RandomSolution::get_parameters() {
- return this->optimal_parameters;
+ std::vector<double>* RandomSolution::get_parameters() {
+ return &this->optimal_parameters;
}
void RandomSolution::optimize(lib4neuro::ErrorFunction &ef, std::ofstream *ofs) {
ef.get_network_instance()->randomize_parameters();
- *this->optimal_parameters = *ef.get_parameters();
- COUT_INFO("Producing a random solution... error: " << ef.eval(this->optimal_parameters) << std::endl);
+ this->optimal_parameters = *ef.get_parameters();
+ COUT_INFO("Producing a random solution... error: " << ef.eval(&this->optimal_parameters) << std::endl);
}
}
\ No newline at end of file
diff --git a/src/LearningMethods/RandomSolution.h b/src/LearningMethods/RandomSolution.h
index 5a18b9dcd6464679ed80d8b301325236178b49a0..38fbdfdfdf982abd989a0ac6022c7c2682cbf1c1 100644
--- a/src/LearningMethods/RandomSolution.h
+++ b/src/LearningMethods/RandomSolution.h
@@ -17,7 +17,7 @@ namespace lib4neuro {
class RandomSolution : public lib4neuro::LearningMethod {
private:
- std::vector<double> *optimal_parameters;
+ std::vector<double> optimal_parameters;
protected:
public: