remove most of the destroy/free_vector calls

that should fix compilation
shogun-toolbox · May 3, 2012 · e2748d0 · e2748d0
1 parent 9d458c8
commit e2748d0
Show file tree

Hide file tree

Showing 24 changed files with 17 additions and 93 deletions.
diff --git a/src/shogun/classifier/AveragedPerceptron.cpp b/src/shogun/classifier/AveragedPerceptron.cpp
@@ -94,7 +94,6 @@ bool CAveragedPerceptron::train(CFeatures* data)
 	bias=tmp_bias/(num_vec*iter);
 
 	SG_FREE(output);
-	train_labels.free_vector();
 	SG_FREE(tmp_w);
 
 	return converged;

diff --git a/src/shogun/classifier/GaussianNaiveBayes.cpp b/src/shogun/classifier/GaussianNaiveBayes.cpp
@@ -42,9 +42,7 @@ CGaussianNaiveBayes::~CGaussianNaiveBayes()
 	SG_UNREF(m_features);
 
 	m_means.destroy_matrix();
-	m_rates.destroy_vector();
 	m_variances.destroy_matrix();
-	m_label_prob.destroy_vector();
 };
 
 CFeatures* CGaussianNaiveBayes::get_features()
@@ -107,18 +105,14 @@ bool CGaussianNaiveBayes::train(CFeatures* data)
 	m_variances.num_rows = m_dim;
 	m_variances.num_cols = m_num_classes;
 
-	m_label_prob.vector = SG_MALLOC(float64_t, m_num_classes);
-	m_label_prob.vlen = m_num_classes;
+	m_label_prob=SGVector<float64_t>(m_num_classes);
 
 	// allocate memory for label rates
-	m_rates.vector = SG_MALLOC(float64_t, m_num_classes);
-	m_rates.vlen = m_num_classes;
+	m_rates=SGVector<float64_t>(m_num_classes);
 
 	// assure that memory is allocated
 	ASSERT(m_means.matrix);
 	ASSERT(m_variances.matrix);
-	ASSERT(m_rates.vector);
-	ASSERT(m_label_prob.vector);
 
 	// make arrays filled by zeros before using
 	m_means.zero();
@@ -139,7 +133,6 @@ bool CGaussianNaiveBayes::train(CFeatures* data)
 		SGVector<float64_t> fea = m_features->get_computed_dot_feature_vector(i);
 		for (j=0; j<m_dim; j++)
 			m_means(j, train_labels.vector[i]) += fea.vector[j];
-		fea.free_vector();
 
 		m_label_prob.vector[train_labels.vector[i]]+=1.0;
 
@@ -166,7 +159,6 @@ bool CGaussianNaiveBayes::train(CFeatures* data)
 			m_variances(j, train_labels.vector[i]) += 
 				CMath::sq(fea[j]-m_means(j, train_labels.vector[i]));
 		}
-		fea.free_vector();
 
 		progress++;
 		SG_PROGRESS(progress, 0, max_progress);
@@ -186,8 +178,6 @@ bool CGaussianNaiveBayes::train(CFeatures* data)
 	}
 	SG_DONE();
 
-	train_labels.free_vector();
-
 	return true;
 }
 
@@ -256,7 +246,6 @@ float64_t CGaussianNaiveBayes::apply(int32_t idx)
 		if (m_rates.vector[i]>m_rates.vector[max_label_idx])
 			max_label_idx = i;
 	}
-	feature_vector.free_vector();
 
 	return max_label_idx+m_min_label;
 };
diff --git a/src/shogun/classifier/LDA.cpp b/src/shogun/classifier/LDA.cpp
@@ -188,7 +188,6 @@ bool CLDA::train_machine(CFeatures* data)
 	CMath::display_vector(mean_neg, num_feat, "mean_neg");
 #endif
 
-	train_labels.free_vector();
 	SG_FREE(mean_neg);
 	SG_FREE(mean_pos);
 	SG_FREE(scatter);

diff --git a/src/shogun/classifier/Perceptron.cpp b/src/shogun/classifier/Perceptron.cpp
@@ -83,7 +83,6 @@ bool CPerceptron::train_machine(CFeatures* data)
 		SG_WARNING("Perceptron algorithm did not converge after %d iterations.\n", max_iter);
 
 	SG_FREE(output);
-	train_labels.free_vector();
 
 	return converged;
 }
diff --git a/src/shogun/classifier/QDA.cpp b/src/shogun/classifier/QDA.cpp
@@ -62,7 +62,6 @@ void CQDA::cleanup()
 	m_covs.free_ndarray();
 	m_M.free_ndarray();
 	m_means.free_matrix();
-	m_slog.free_vector();
 
 	m_num_classes = 0;
 }
@@ -81,7 +80,6 @@ CLabels* CQDA::apply()
 	SGMatrix< float64_t > X(num_vecs, m_dim);
 	SGMatrix< float64_t > A(num_vecs, m_dim);
 	SGVector< float64_t > norm2(num_vecs*m_num_classes);
-
 	norm2.zero();
 
 	int i, j, k, vlen;
@@ -132,7 +130,6 @@ CLabels* CQDA::apply()
 	CMath::display_vector(out->get_labels().vector, num_vecs, "Labels");
 #endif
 
-	norm2.destroy_vector();
 	A.destroy_matrix();
 	X.destroy_matrix();
 
@@ -300,7 +297,7 @@ bool CQDA::train_machine(CFeatures* data)
 	M_dims[2] = m_num_classes;
 	m_M = SGNDArray< float64_t >(M_dims, 3, true);
 
-	m_slog = SGVector< float32_t >(m_num_classes, true);
+	m_slog = SGVector< float32_t >(m_num_classes);
 	m_slog.zero();
 
 	index_t idx = 0;
@@ -345,8 +342,6 @@ bool CQDA::train_machine(CFeatures* data)
 
 	rotations.destroy_ndarray();
 	scalings.destroy_matrix();
-	sinvsqrt.destroy_vector();
-	train_labels.destroy_vector();
 	SG_FREE(class_idxs);
 	SG_FREE(class_nums);
 	return true;

diff --git a/src/shogun/classifier/mkl/MKL.cpp b/src/shogun/classifier/mkl/MKL.cpp
@@ -1473,7 +1473,7 @@ void CMKL::compute_sum_beta(float64_t* sumw)
 
 	int32_t nsv=svm->get_num_support_vectors();
 	int32_t num_kernels = kernel->get_num_subkernels();
-	SGVector<float64_t> beta=SGVector<float64_t>(num_kernels, true);
+	SGVector<float64_t> beta=SGVector<float64_t>(num_kernels);
 	int32_t nweights=0;
 	const float64_t* old_beta = kernel->get_subkernel_weights(nweights);
 	ASSERT(nweights==num_kernels);
@@ -1507,9 +1507,6 @@ void CMKL::compute_sum_beta(float64_t* sumw)
 
 	mkl_iterations++;
 	kernel->set_subkernel_weights(SGVector<float64_t>( (float64_t*) old_beta, num_kernels));
-
-	/* safe because of above comment, otherwise: memleak */
-	beta.free_vector();
 }
 
 

diff --git a/src/shogun/classifier/mkl/MKLMulticlass.cpp b/src/shogun/classifier/mkl/MKLMulticlass.cpp
@@ -241,11 +241,9 @@ void CMKLMulticlass::addingweightsstep( const std::vector<float64_t> &
 
 float64_t CMKLMulticlass::getsumofsignfreealphas()
 {
-
 	std::vector<int> trainlabels2(m_labels->get_num_labels());
 	SGVector<int32_t> lab=m_labels->get_int_labels();
 	std::copy(lab.vector,lab.vector+lab.vlen, trainlabels2.begin());
-	lab.free_vector();
 
 	ASSERT (trainlabels2.size()>0);
 	float64_t sum=0;

diff --git a/src/shogun/classifier/svm/NewtonSVM.cpp b/src/shogun/classifier/svm/NewtonSVM.cpp
@@ -108,7 +108,6 @@ bool CNewtonSVM::train_machine(CFeatures* data)
 			sgv=features->get_computed_dot_feature_vector(sv[k]);
 			for (int32_t j=0; j<x_d; j++)
 				Xsv[k*x_d+j]=sgv.vector[j];
-			sgv.destroy_vector();
 		}
 		int32_t tx=x_d;
 		int32_t ty=size_sv;

diff --git a/src/shogun/classifier/svm/SGDQN.cpp b/src/shogun/classifier/svm/SGDQN.cpp
@@ -196,8 +196,6 @@ bool CSGDQN::train(CFeatures* data)
 				}
 			}
 			t++;
-
-			v.free_vector();
 		}
 	}
 	SG_FREE(result);

diff --git a/src/shogun/classifier/svm/SVM.cpp b/src/shogun/classifier/svm/SVM.cpp
@@ -41,7 +41,6 @@ CSVM::CSVM(float64_t C, CKernel* k, CLabels* lab)
 
 CSVM::~CSVM()
 {
-	m_linear_term.destroy_vector();
 	SG_UNREF(mkl);
 }
 
@@ -303,15 +302,15 @@ float64_t* CSVM::get_linear_term_array()
 {
 	if (m_linear_term.vlen==0)
 		return NULL;
+	float64_t* a = SG_MALLOC(float64_t, m_linear_term.vlen);
 
-	SGVector<float64_t> a(m_linear_term.vlen);
-	memcpy(a.vector, m_linear_term.vector,
+	memcpy(a, m_linear_term.vector,
 			m_linear_term.vlen*sizeof(float64_t));
 
-	return a.vector;
+	return a;
 }
 
-void CSVM::set_linear_term(const SGVector<float64_t>& linear_term)
+void CSVM::set_linear_term(const SGVector<float64_t> linear_term)
 {
 	ASSERT(linear_term.vector);
 
@@ -326,12 +325,7 @@ void CSVM::set_linear_term(const SGVector<float64_t>& linear_term)
 				"of entries (%d) in linear term \n", num_labels, linear_term.vlen);
 	}
 
-	m_linear_term.destroy_vector();
-
-	m_linear_term.vlen=linear_term.vlen;
-	m_linear_term=SGVector<float64_t> (linear_term.vlen);
-	memcpy(m_linear_term.vector, linear_term.vector,
-			linear_term.vlen*sizeof(float64_t));
+	m_linear_term=linear_term;
 }
 
 SGVector<float64_t> CSVM::get_linear_term()

diff --git a/src/shogun/classifier/svm/SVM.h b/src/shogun/classifier/svm/SVM.h
@@ -81,7 +81,7 @@ class CSVM : public CKernelMachine
 		 *
 		 * @param linear_term the linear term
 		 */
-		virtual void set_linear_term(const SGVector<float64_t>& linear_term);
+		virtual void set_linear_term(const SGVector<float64_t> linear_term);
 
 
 		/** load a SVM from file

diff --git a/src/shogun/classifier/svm/SVMLin.cpp b/src/shogun/classifier/svm/SVMLin.cpp
@@ -105,6 +105,5 @@ bool CSVMLin::train_machine(CFeatures* data)
 
 	SG_FREE(Data.C);
 	SG_FREE(Outputs.vec);
-	train_labels.free_vector();
 	return true;
 }
diff --git a/src/shogun/classifier/svm/SVMOcas.cpp b/src/shogun/classifier/svm/SVMOcas.cpp
@@ -135,8 +135,6 @@ bool CSVMOcas::train_machine(CFeatures* data)
 	SG_FREE(cp_bias);
 	cp_bias=NULL;
 
-	lab.free_vector();
-
 	SG_FREE(old_w);
 	old_w=NULL;
 
@@ -340,7 +338,6 @@ void CSVMOcas::init()
 	w=NULL;
 	old_w=NULL;
 	tmp_a_buf=NULL;
-	lab.destroy_vector();
 	cp_value=NULL;
 	cp_index=NULL;
 	cp_nz_dims=NULL;

diff --git a/src/shogun/classifier/svm/WDSVMOcas.cpp b/src/shogun/classifier/svm/WDSVMOcas.cpp
@@ -244,8 +244,6 @@ bool CWDSVMOcas::train_machine(CFeatures* data)
 	SG_FREE(cuts);
 
 	lab=NULL;
-	labvec.free_vector();
-
 	SG_UNREF(alphabet);
 
 	return true;

diff --git a/src/shogun/clustering/GMM.cpp b/src/shogun/clustering/GMM.cpp
@@ -106,8 +106,8 @@ void CGMM::cleanup()
 	for (int32_t i = 0; i < m_components.vlen; i++)
 		SG_UNREF(m_components.vector[i]);
 
-	m_components.destroy_vector();
-	m_coefficients.destroy_vector();
+	m_components.unref();
+	m_coefficients.unref();
 }
 
 bool CGMM::train(CFeatures* data)
@@ -178,7 +178,6 @@ float64_t CGMM::train_em(float64_t min_cov, int32_t max_iter, float64_t min_chan
 
 			logPx[i]=CMath::log(logPx[i]);
 			log_likelihood_cur+=logPx[i];
-			v.free_vector();
 
 			for (int32_t j=0; j<m_components.vlen; j++)
 			{
@@ -244,7 +243,6 @@ float64_t CGMM::train_smem(int32_t max_iter, int32_t max_cand, float64_t min_cov
 			}
 
 			logPx[i]=CMath::log(logPx[i]);
-			v.free_vector();
 
 			for (int32_t j=0; j<m_components.vlen; j++)
 			{
@@ -372,7 +370,6 @@ void CGMM::partial_em(int32_t comp1, int32_t comp2, int32_t comp3, float64_t min
 		post_add[i]=CMath::log(CMath::exp(init_logPxy[i*m_components.vlen+comp1]-init_logPx[i])+
 					CMath::exp(init_logPxy[i*m_components.vlen+comp2]-init_logPx[i])+
 					CMath::exp(init_logPxy[i*m_components.vlen+comp3]-init_logPx[i]));
-		v.free_vector();
 	}
 
 	SGVector<CGaussian*> components(3);
@@ -495,7 +492,6 @@ void CGMM::partial_em(int32_t comp1, int32_t comp2, int32_t comp3, float64_t min
 
 			logPx[i]=CMath::log(logPx[i]+init_logPx_fix[i]);
 			log_likelihood_cur+=logPx[i];
-			v.free_vector();
 
 			for (int32_t j=0; j<3; j++)
 			{
@@ -549,7 +545,6 @@ void CGMM::max_likelihood(SGMatrix<float64_t> alpha, float64_t min_cov)
 			alpha_sum+=alpha.matrix[j*alpha.num_cols+i];
 			SGVector<float64_t> v=dotdata->get_computed_dot_feature_vector(j);
 			CMath::add<float64_t>(mean_sum, alpha.matrix[j*alpha.num_cols+i], v.vector, 1, mean_sum, v.vlen);
-			v.free_vector();
 		}
 
 		for (int32_t j=0; j<num_dim; j++)
@@ -601,8 +596,6 @@ void CGMM::max_likelihood(SGMatrix<float64_t> alpha, float64_t min_cov)
 					cov_sum[0]+=temp*alpha.matrix[j*alpha.num_cols+i];
 					break;
 			}
-
-			v.free_vector();
 		}
 
 		switch (cov_type)
@@ -705,9 +698,8 @@ SGVector<float64_t> CGMM::get_coef()
 	return m_coefficients;
 }
 
-void CGMM::set_coef(const SGVector<float64_t>& coefficients)
+void CGMM::set_coef(const SGVector<float64_t> coefficients)
 {
-	m_coefficients.destroy_vector();
 	m_coefficients=coefficients;
 }
 
@@ -723,7 +715,6 @@ void CGMM::set_comp(const SGVector<CGaussian*>& components)
 		SG_UNREF(m_components.vector[i]);
 	}
 
-	m_components.destroy_vector();
 	m_components=components;
 
 	for (int32_t i=0; i<m_components.vlen; i++)

diff --git a/src/shogun/clustering/GMM.h b/src/shogun/clustering/GMM.h
@@ -177,7 +177,7 @@ class CGMM : public CDistribution
 		 *
 		 * @param coefficients mixing coefficients
 		 */
-		virtual void set_coef(const SGVector<float64_t>& coefficients);
+		virtual void set_coef(const SGVector<float64_t> coefficients);
 
 		/** get components
 		 *

diff --git a/src/shogun/clustering/KMeans.cpp b/src/shogun/clustering/KMeans.cpp
@@ -42,7 +42,6 @@ CKMeans::CKMeans(int32_t k_, CDistance* d)
 
 CKMeans::~CKMeans()
 {
-	R.destroy_vector();
 }
 
 bool CKMeans::train_machine(CFeatures* data)
@@ -65,7 +64,6 @@ bool CKMeans::train_machine(CFeatures* data)
 		Weights.vector[i]=1.0;
 
 	clustknb(false, NULL);
-	Weights.destroy_vector();
 
 	return true;
 }
@@ -188,7 +186,6 @@ void CKMeans::clustknb(bool use_old_mus, float64_t *mus_start)
 	const int32_t XDimk=dimensions*k;
 	int32_t iter=0;
 
-	R.destroy_vector();
 	R=SGVector<float64_t>(k);
 
 	mus=SGMatrix<float64_t>(dimensions, k);

diff --git a/src/shogun/distance/MahalanobisDistance.cpp b/src/shogun/distance/MahalanobisDistance.cpp
@@ -91,9 +91,6 @@ float64_t CMahalanobisDistance::compute(int32_t idx_a, int32_t idx_b)
 
 	float64_t result = cblas_ddot(v.vlen, v.vector, 1, diff.vector, 1);
 
-	v.destroy_vector();
-	diff.destroy_vector();
-
 	if (!use_mean)
 		((CSimpleFeatures<float64_t>*) lhs)->free_feature_vector(avec, idx_a);