fixed some python_modular warnings

src/shogun/machine/gp/InferenceMethod.h

@@ -181,7 +181,7 @@ class CInferenceMethod : public CDifferentiableFunction
 	 *
 	 * where \f$\mu\f$ is the mean,
 	 * \f$K\f$ is the prior covariance matrix,
-	 * and \f$meanf$\f is the mean prior fomr MeanFunction
+	 * and \f$meanf\f$ is the mean prior fomr MeanFunction
 	 *
 	 */
 	virtual SGVector<float64_t> get_alpha()=0;

src/shogun/machine/gp/KLCovarianceInferenceMethod.h

@@ -179,7 +179,7 @@ class CKLCovarianceInferenceMethod: public CKLInferenceMethod
 	 * get_derivative_wrt_kernel(const TParameter* param)
 	 * will call this function
 	 *
-	 * @param the gradient wrt hyperparameter related to cov
+	 * @param dK the gradient wrt hyperparameter related to cov
 	 */
 
 	virtual float64_t get_derivative_related_cov(SGMatrix<float64_t> dK);

src/shogun/machine/gp/KLDualInferenceMethod.h

@@ -189,7 +189,7 @@ class CKLDualInferenceMethod: public CKLInferenceMethod
 	 * get_derivative_wrt_kernel(const TParameter* param)
 	 * will call this function
 	 *
-	 * @param the gradient related to cov
+	 * @param dK the gradient related to cov
 	 *
 	 * @return the gradient wrt hyperparameter related to cov
 	 */

src/shogun/machine/gp/KLInferenceMethod.h

@@ -332,7 +332,7 @@ class CKLInferenceMethod: public CInferenceMethod
 	 * get_derivative_wrt_kernel(const TParameter* param)
 	 * will call this function
 	 *
-	 * @param the gradient wrt hyperparameter related to cov
+	 * @param dK the gradient wrt hyperparameter related to cov
 	 */
 	virtual float64_t get_derivative_related_cov(SGMatrix<float64_t> dK)=0;
 

src/shogun/machine/gp/KLLowerTriangularInferenceMethod.h

@@ -115,7 +115,7 @@ class CKLLowerTriangularInferenceMethod: public CKLInferenceMethod
 	 * get_derivative_wrt_kernel(const TParameter* param)
 	 * will call this function
 	 *
-	 * @param the gradient wrt hyperparameter related to cov
+	 * @param dK the gradient wrt hyperparameter related to cov
 	 */
 
 	virtual float64_t get_derivative_related_cov(SGMatrix<float64_t> dK);

src/shogun/machine/gp/LaplacianInferenceBase.h

@@ -89,7 +89,7 @@ class CLaplacianInferenceBase: public CInferenceMethod
 	 *
 	 * where \f$\mu\f$ is the mean,
 	 * \f$K\f$ is the prior covariance matrix,
-	 * and \f$meanf$\f is the mean prior fomr MeanFunction
+	 * and \f$meanf\f$ is the mean prior fomr MeanFunction
 	 *
 	 */
 	virtual SGVector<float64_t> get_alpha();
@@ -98,23 +98,23 @@ class CLaplacianInferenceBase: public CInferenceMethod
 	 *
 	 * @return Cholesky decomposition of matrix:
 	 *
-	 *
 	 * for binary and regression case
 	 * \f[
 	 * L = Cholesky(W^{\frac{1}{2}}*K*W^{\frac{1}{2}}+I)
 	 * \f]
-	 *
+
 	 * where \f$K\f$ is the prior covariance matrix, \f$sW\f$ is the vector
-	 * returned by get_diagonal_vector(), and \f$I\f$ is the identity matrix.
-	 *
-	 * for multiclass case
-	 * \f[
-	 * M = Cholesky(\sum_\text{c}{E_\text{c})
-	 * \f]
+         * returned by get_diagonal_vector(), and \f$I\f$ is the identity matrix.
+         *
+         * for multiclass case
+         * \f[
+         * M = Cholesky(\sum_\text{c}{E_\text{c})
+         * \f]
+         *
+         * where \f$E_\text{c}\f$ is the matrix defined in the algorithm 3.3 of the GPML textbook for class c
+         * Note the E matrix is used to store these \f$E_\text{c}\f$ matrices, where E=[E_1, E_2, ..., E_C],
+         * where C is the number of classes and C should be greater than 1.
 	 *
-	 * where \f$E_\text{c}\f$ is the matrix defined in the algorithm 3.3 of the GPML textbook for class c
-	 * Note the E matrix is used to store these \f$E_\text{c}\f$ matrices, where E=[E_1, E_2, ..., E_C],
-	 * where C is the number of classes and C should be greater than 1.
 	 */
 	virtual SGMatrix<float64_t> get_cholesky();
 

src/shogun/machine/gp/LogitDVGLikelihood.h

@@ -53,7 +53,7 @@ namespace shogun
  *
  * The mathematically definition (equation 19 in the paper) is as below
  * \f[
- * Fenchel_i(\alpha_i,\lambda_i) = max_{h_i,\rho_i}{\alpha_i h_i+\lambda_i \rho_i /2 - E_{q(f_i|h_i,\rho_i)}(-log(p(y_i|f_i)))}
+ * \text{Fenchel}_i(\alpha_i,\lambda_i) = \max_{h_i,\rho_i}{\alpha_i h_i+\lambda_i \rho_i /2 - E_{q(f_i|h_i,\rho_i)}(-log(p(y_i|f_i)))}
  * \f]
  * where \f$\alpha_i\f$,\f$\lambda_i\f$ are Lagrange multipliers with respective to constraints
  * \f$h_i=\mu_i\f$ and \f$\rho_i=\sigma_i^2\f$ respectively,
@@ -63,7 +63,7 @@ namespace shogun
  * In this setting, \f$\alpha\f$ and \f$\lambda\f$ are called dual parameters for \f$\mu\f$ and \f$\sigma^2\f$ respectively.
  *
  * Note that \f$p(y_i)\f$ is Logistic distribution and a local variational bound defined as below is used to approximate
- * -E_{q(f_i|h_i,\rho_i)}(-log(p(y_i|f_i)))
+ * \f$-\text{E}_{q(f_i|h_i,\rho_i)}(-\log(p(y_i|f_i)))\f$
  *
  * The local variational bound used here is
  * \f[

src/shogun/machine/gp/MultiLaplacianInferenceMethod.h

@@ -120,7 +120,7 @@ class CMultiLaplacianInferenceMethod: public CLaplacianInferenceBase
 	virtual float64_t get_negative_log_marginal_likelihood();
 
 	/** get diagonal vector
-	 * where the vector, \f$\pi$\f, defined in the algorithm 3.3 of the GPML textbook
+	 * where the vector, \f$\pi\f$, defined in the algorithm 3.3 of the GPML textbook
 	 *
 	 * @return the vector used for inference
 	 */
@@ -221,7 +221,7 @@ class CMultiLaplacianInferenceMethod: public CLaplacianInferenceBase
 
 	/** the helper method used to compute gradient of GP wrt hyperparameter
 	 *
-	 * @param raw gradient wrt hyperparameter
+	 * @param dK raw gradient wrt hyperparameter
 	 *
 	 * @return the gradient of GP wrt hyperparameter
 	 *

src/shogun/machine/gp/SingleFITCLaplacianInferenceMethod.h

@@ -289,7 +289,7 @@ friend class CFITCPsiLine;
 	virtual SGMatrix<float64_t> get_chol_inv(SGMatrix<float64_t> mtx);
 
 	/** efficiently compute the matrix-vector product
-	 * \f$\Sigma \times al$\f, where \f$\Sigma$\f is
+	 * \f$\Sigma \times al\f$, where \f$\Sigma\f$ is
 	 * the FITC equivalent covariance n-by-n matrix (prior) of f_n
 	 *
 	 * @param al input vector
@@ -299,7 +299,7 @@ friend class CFITCPsiLine;
 	virtual SGVector<float64_t> compute_mvmK(SGVector<float64_t> al);
 
 	/** efficiently compute the matrix-vector product
-	 * \f$ \inv{\inv{W}+\Sigma} \times x$\f, where \f$\Sigma$\f is
+	 * \f$ \inv{\inv{W}+\Sigma} \times x\f$, where \f$\Sigma\f$ is
 	 * the FITC equivalent covariance n-by-n matrix (prior) of f_n
 	 *
 	 * @param x input vector

src/shogun/machine/gp/SingleFITCLaplacianInferenceMethodWithLBFGS.h

@@ -60,7 +60,7 @@ class CSingleFITCLaplacianInferenceMethodWithLBFGS: public CSingleFITCLaplacianI
 	/* default constructor */
 	CSingleFITCLaplacianInferenceMethodWithLBFGS();
 
-	/* constructor
+	/** constructor
 	 *
 	 * @param kernel covariance function
 	 * @param features features to use in inference
@@ -86,7 +86,7 @@ class CSingleFITCLaplacianInferenceMethodWithLBFGS: public CSingleFITCLaplacianI
 	virtual const char* get_name() const
 	{return "SingleFITCLaplacianInferenceMethodWithLBFGS";}
 
-	/* set L-BFGS parameters
+	/** set L-BFGS parameters
 	 * For details please see shogun/optimization/lbfgs/lbfgs.h
 	 * @param m The number of corrections to approximate the inverse hessian matrix.
 	 * Default value is 100.
@@ -150,60 +150,60 @@ class CSingleFITCLaplacianInferenceMethodWithLBFGS: public CSingleFITCLaplacianI
 	virtual void update_alpha();
 
 private:
-	/* a parameter used to compute function value and gradient for LBFGS update*/
+	/** a parameter used to compute function value and gradient for LBFGS update*/
 	SGVector<float64_t> * m_mean_f;
 
-	/* should we enable the original Newton method
+	/** should we enable the original Newton method
 	 * if the L-BFGS method fails
 	 * */
 	bool m_enable_newton_if_fail;
 
-	/* The number of corrections to approximate the inverse hessian matrix.*/
+	/** The number of corrections to approximate the inverse hessian matrix.*/
 	int m_m;
 
-	/* The maximum number of trials to do line search for each L-BFGS update.*/
+	/** The maximum number of trials to do line search for each L-BFGS update.*/
 	int m_max_linesearch;
 
-	/* The line search algorithm.*/
+	/** The line search algorithm.*/
 	int m_linesearch;
 
-	/* The maximum number of iterations for L-BFGS update.*/
+	/** The maximum number of iterations for L-BFGS update.*/
 	int m_max_iterations;
 
-	/* Delta for convergence test based on the change of function value.*/
+	/** Delta for convergence test based on the change of function value.*/
 	float64_t m_delta;
 
-	/* Distance for delta-based convergence test.*/
+	/** Distance for delta-based convergence test.*/
 	int m_past;
 
-	/* Epsilon for convergence test based on the change of gradient.*/
+	/** Epsilon for convergence test based on the change of gradient.*/
 	float64_t m_epsilon;
 
-	/* The minimum step of the line search.*/
+	/** The minimum step of the line search.*/
 	float64_t m_min_step;
 
-	/* The maximum step of the line search.*/
+	/** The maximum step of the line search.*/
 	float64_t m_max_step;
 
-	/* A parameter used in Armijo condition.*/
+	/** A parameter used in Armijo condition.*/
 	float64_t m_ftol;
 
-	/* A parameter used in curvature condition.*/
+	/** A parameter used in curvature condition.*/
 	float64_t m_wolfe;
 
-	/* A parameter used in Morethuente linesearch to control the accuracy.*/
+	/** A parameter used in Morethuente linesearch to control the accuracy.*/
 	float64_t m_gtol;
 
-	/* The machine precision for floating-point values.*/
+	/** The machine precision for floating-point values.*/
 	float64_t m_xtol;
 
-	/* Coeefficient for the L1 norm of variables.*/
+	/** Coeefficient for the L1 norm of variables.*/
 	float64_t m_orthantwise_c;
 
-	/* Start index for computing L1 norm of the variables.*/
+	/** Start index for computing L1 norm of the variables.*/
 	int m_orthantwise_start;
 
-	/* End index for computing L1 norm of the variables.*/
+	/** End index for computing L1 norm of the variables.*/
 	int m_orthantwise_end;
 
 	void init();

src/shogun/machine/gp/SingleLaplacianInferenceMethodWithLBFGS.h

@@ -90,7 +90,7 @@ class CSingleLaplacianInferenceMethodWithLBFGS: public CSingleLaplacianInference
 	virtual const char* get_name() const
 	{return "SingleLaplacianInferenceMethodWithLBFGS";}
 
-	/* set L-BFGS parameters
+	/** set L-BFGS parameters
 	 * For details please see shogun/optimization/lbfgs/lbfgs.h
 	 * @param m The number of corrections to approximate the inverse hessian matrix.
 	 * Default value is 100.
@@ -144,7 +144,7 @@ class CSingleLaplacianInferenceMethodWithLBFGS: public CSingleLaplacianInference
 			int orthantwise_start = 0,
 			int orthantwise_end = 1);
 
-	/* wheter we use Newton method as rollbak if LBFGS optimizer fails
+	/** wheter we use Newton method as rollbak if LBFGS optimizer fails
 	 *
 	 * @param enable_newton_if_fail if LBFGS optimizer fails, should we use Newton method.
 	 */
@@ -154,60 +154,60 @@ class CSingleLaplacianInferenceMethodWithLBFGS: public CSingleLaplacianInference
 	virtual void update_alpha();
 
 private:
-	/* a parameter used to compute function value and gradient for LBFGS update*/
+	/** a parameter used to compute function value and gradient for LBFGS update*/
 	SGVector<float64_t> * m_mean_f;
 
-	/* should we enable the original Newton method
+	/** should we enable the original Newton method
 	 * if the L-BFGS method fails
 	 * */
 	bool m_enable_newton_if_fail;
 
-	/* The number of corrections to approximate the inverse hessian matrix.*/
+	/** The number of corrections to approximate the inverse hessian matrix.*/
 	int m_m;
 
-	/* The maximum number of trials to do line search for each L-BFGS update.*/
+	/** The maximum number of trials to do line search for each L-BFGS update.*/
 	int m_max_linesearch;
 
-	/* The line search algorithm.*/
+	/** The line search algorithm.*/
 	int m_linesearch;
 
-	/* The maximum number of iterations for L-BFGS update.*/
+	/** The maximum number of iterations for L-BFGS update.*/
 	int m_max_iterations;
 
-	/* Delta for convergence test based on the change of function value.*/
+	/** Delta for convergence test based on the change of function value.*/
 	float64_t m_delta;
 
-	/* Distance for delta-based convergence test.*/
+	/** Distance for delta-based convergence test.*/
 	int m_past;
 
-	/* Epsilon for convergence test based on the change of gradient.*/
+	/** Epsilon for convergence test based on the change of gradient.*/
 	float64_t m_epsilon;
 
-	/* The minimum step of the line search.*/
+	/** The minimum step of the line search.*/
 	float64_t m_min_step;
 
-	/* The maximum step of the line search.*/
+	/** The maximum step of the line search.*/
 	float64_t m_max_step;
 
-	/* A parameter used in Armijo condition.*/
+	/** A parameter used in Armijo condition.*/
 	float64_t m_ftol;
 
-	/* A parameter used in curvature condition.*/
+	/** A parameter used in curvature condition.*/
 	float64_t m_wolfe;
 
-	/* A parameter used in Morethuente linesearch to control the accuracy.*/
+	/** A parameter used in Morethuente linesearch to control the accuracy.*/
 	float64_t m_gtol;
 
-	/* The machine precision for floating-point values.*/
+	/** The machine precision for floating-point values.*/
 	float64_t m_xtol;
 
-	/* Coeefficient for the L1 norm of variables.*/
+	/** Coeefficient for the L1 norm of variables.*/
 	float64_t m_orthantwise_c;
 
-	/* Start index for computing L1 norm of the variables.*/
+	/** Start index for computing L1 norm of the variables.*/
 	int m_orthantwise_start;
 
-	/* End index for computing L1 norm of the variables.*/
+	/** End index for computing L1 norm of the variables.*/
 	int m_orthantwise_end;
 
 	void init();

src/shogun/machine/gp/SoftMaxLikelihood.h

@@ -60,7 +60,7 @@ enum EMCSamplerType
 
 /** @brief Class that models Soft-Max likelihood.
  *
- * softmax_i(f)=\frac{\exp{f_i}}{\sum\exp{f_i}}
+ * \f$ \text{softmax}_i(f)=\frac{\exp{f_i}}{\sum\exp{f_i}} \f$
  *
  * Code adapted from
  * https://gist.github.com/yorkerlin/8a36e8f9b298aa0246a4
@@ -285,13 +285,13 @@ class CSoftMaxLikelihood : public CLikelihoodModel
 	SGVector<float64_t> predictive_helper(SGVector<float64_t> mu,
 	SGVector<float64_t> s2, const CLabels *lab, EMCSamplerType option) const;
 
-	/**the Monte method sampler
+	/** the Monte method sampler
 	 *
 	 * @param num_samples number of samples to be generated
-	 * @param mu posterior mean of a Gaussian distribution
+	 * @param mean posterior mean of a Gaussian distribution
 	 * \f$\mathcal{N}(\mu,\sigma^2)\f$, which is an approximation to the
 	 * posterior marginal \f$p(f_*|X,y,x_*)\f$
-	 * @param s2 posterior variance of a Gaussian distribution
+	 * @param Sigma posterior variance of a Gaussian distribution
 	 * \f$\mathcal{N}(\mu,\sigma^2)\f$, which is an approximation to the
 	 * posterior marginal \f$p(f_*|X,y,x_*)\f$
 	 * @param y labels based on 0 and 1 encoding \f$y_*\f$

src/shogun/machine/gp/SparseInferenceBase.h

@@ -288,7 +288,7 @@ class CSparseInferenceBase: public CInferenceMethod
 	 * where
 	 * \f$\lambda \ge 0\f$ is the inducing noise.
 	 *
-	 * In practice, we use the corrected matrix, \Sigma_{M'} in the following approximation.
+	 * In practice, we use the corrected matrix, \f$\Sigma_{M'}\f$ in the following approximation.
 	 *\f[
 	 *\Sigma_{Sparse}=\textbf{diag}(\Sigma_{N}-\Phi)+\Phi
 	 *\f]

src/shogun/mathematics/Integration.h

@@ -114,8 +114,8 @@ class CIntegration : public CSGObject
 	 * Gauss-Hermite formula respectively.
 	 *
 	 * @param f integrable function of one variable
-	 * @param x the provided array of nodes
-	 * @param w the provided array of weights
+	 * @param xgh the provided array of nodes
+	 * @param wgh the provided array of weights
 	 *
 	 * @return approximate value of the
 	 * integral \f$\int_{-\infty}^{\infty}e^{-x^2}f(x)dx\f$