// SPDX-License-Identifier: Apache-2.0
// 
// Copyright 2008-2016 Conrad Sanderson (https://conradsanderson.id.au)
// Copyright 2008-2016 National ICT Australia (NICTA)
// 
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// https://www.apache.org/licenses/LICENSE-2.0
// 
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
// ------------------------------------------------------------------------


//! \addtogroup op_princomp
//! @{



//! \brief
//! principal component analysis -- 4 arguments version
//! computation is done via singular value decomposition
//! coeff_out    -> principal component coefficients
//! score_out    -> projected samples
//! latent_out   -> eigenvalues of principal vectors
//! tsquared_out -> Hotelling's T^2 statistic
template<typename T1>
inline
bool
op_princomp::direct_princomp
  (
         Mat<typename T1::elem_type>&     coeff_out,
         Mat<typename T1::elem_type>&     score_out,
         Col<typename T1::pod_type>&      latent_out,
         Col<typename T1::elem_type>&     tsquared_out,
  const Base<typename T1::elem_type, T1>& X
  )
  {
  arma_debug_sigprint();
  
  typedef typename T1::elem_type eT;
  typedef typename T1::pod_type   T;
  
  const unwrap_check<T1> Y( X.get_ref(), score_out );
  const Mat<eT>& in    = Y.M;

  const uword n_rows = in.n_rows;
  const uword n_cols = in.n_cols;
  
  if(n_rows > 1) // more than one sample
    {
    // subtract the mean - use score_out as temporary matrix
    score_out = in;  score_out.each_row() -= mean(in);
    
    // singular value decomposition
    Mat<eT> U;
    Col< T> s;
    
    const bool svd_ok = (n_rows >= n_cols) ? svd_econ(U, s, coeff_out, score_out) : svd(U, s, coeff_out, score_out);
    
    if(svd_ok == false)  { return false; }
    
    // normalize the eigenvalues
    s /= std::sqrt( double(n_rows - 1) );
    
    // project the samples to the principals
    score_out *= coeff_out;
    
    if(n_rows <= n_cols) // number of samples is less than their dimensionality
      {
      score_out.cols(n_rows-1,n_cols-1).zeros();
      
      Col<T> s_tmp(n_cols, arma_zeros_indicator());
      
      s_tmp.rows(0,n_rows-2) = s.rows(0,n_rows-2);
      s = s_tmp;
          
      // compute the Hotelling's T-squared
      s_tmp.rows(0,n_rows-2) = T(1) / s_tmp.rows(0,n_rows-2);
      
      const Mat<eT> S = score_out * diagmat(Col<T>(s_tmp));
      tsquared_out = sum(S%S,1);
      }
    else
      {
      // compute the Hotelling's T-squared
      // TODO: replace with more robust approach
      const Mat<eT> S = score_out * diagmat(Col<T>( T(1) / s));
      tsquared_out = sum(S%S,1);
      }
    
    // compute the eigenvalues of the principal vectors
    latent_out = s%s;
    }
  else // 0 or 1 samples
    {
    coeff_out.eye(n_cols, n_cols);
    
    score_out.copy_size(in);
    score_out.zeros();
    
    latent_out.set_size(n_cols);
    latent_out.zeros();
    
    tsquared_out.set_size(n_rows);
    tsquared_out.zeros();
    }
  
  return true;
  }



//! \brief
//! principal component analysis -- 3 arguments version
//! computation is done via singular value decomposition
//! coeff_out    -> principal component coefficients
//! score_out    -> projected samples
//! latent_out   -> eigenvalues of principal vectors
template<typename T1>
inline
bool
op_princomp::direct_princomp
  (
         Mat<typename T1::elem_type>&     coeff_out,
         Mat<typename T1::elem_type>&     score_out,
         Col<typename T1::pod_type>&      latent_out,
  const Base<typename T1::elem_type, T1>& X
  )
  {
  arma_debug_sigprint();
  
  typedef typename T1::elem_type eT;
  typedef typename T1::pod_type   T;
  
  const unwrap_check<T1> Y( X.get_ref(), score_out );
  const Mat<eT>& in    = Y.M;
  
  const uword n_rows = in.n_rows;
  const uword n_cols = in.n_cols;
  
  if(n_rows > 1) // more than one sample
    {
    // subtract the mean - use score_out as temporary matrix
    score_out = in;  score_out.each_row() -= mean(in);
    
    // singular value decomposition
    Mat<eT> U;
    Col< T> s;
    
    const bool svd_ok = (n_rows >= n_cols) ? svd_econ(U, s, coeff_out, score_out) : svd(U, s, coeff_out, score_out);
    
    if(svd_ok == false)  { return false; }
    
    // normalize the eigenvalues
    s /= std::sqrt( double(n_rows - 1) );
    
    // project the samples to the principals
    score_out *= coeff_out;
    
    if(n_rows <= n_cols) // number of samples is less than their dimensionality
      {
      score_out.cols(n_rows-1,n_cols-1).zeros();
      
      Col<T> s_tmp(n_cols, arma_zeros_indicator());
      
      s_tmp.rows(0,n_rows-2) = s.rows(0,n_rows-2);
      s = s_tmp;
      }
    
    // compute the eigenvalues of the principal vectors
    latent_out = s%s;
    }
  else // 0 or 1 samples
    {
    coeff_out.eye(n_cols, n_cols);
    
    score_out.copy_size(in);
    score_out.zeros();
    
    latent_out.set_size(n_cols);
    latent_out.zeros(); 
    }
  
  return true;
  }



//! \brief
//! principal component analysis -- 2 arguments version
//! computation is done via singular value decomposition
//! coeff_out    -> principal component coefficients
//! score_out    -> projected samples
template<typename T1>
inline
bool
op_princomp::direct_princomp
  (
         Mat<typename T1::elem_type>&     coeff_out,
         Mat<typename T1::elem_type>&     score_out,
  const Base<typename T1::elem_type, T1>& X
  )
  {
  arma_debug_sigprint();
  
  typedef typename T1::elem_type eT;
  typedef typename T1::pod_type   T;
  
  const unwrap_check<T1> Y( X.get_ref(), score_out );
  const Mat<eT>& in    = Y.M;
  
  const uword n_rows = in.n_rows;
  const uword n_cols = in.n_cols;
  
  if(n_rows > 1) // more than one sample
    {
    // subtract the mean - use score_out as temporary matrix
    score_out = in;  score_out.each_row() -= mean(in);
    
    // singular value decomposition
    Mat<eT> U;
    Col< T> s;
    
    const bool svd_ok = (n_rows >= n_cols) ? svd_econ(U, s, coeff_out, score_out) : svd(U, s, coeff_out, score_out);
    
    if(svd_ok == false)  { return false; }
    
    // project the samples to the principals
    score_out *= coeff_out;
    
    if(n_rows <= n_cols) // number of samples is less than their dimensionality
      {
      score_out.cols(n_rows-1,n_cols-1).zeros();
      }
    }
  else // 0 or 1 samples
    {
    coeff_out.eye(n_cols, n_cols);
    score_out.copy_size(in);
    score_out.zeros();
    }
  
  return true;
  }



//! \brief
//! principal component analysis -- 1 argument version
//! computation is done via singular value decomposition
//! coeff_out    -> principal component coefficients
template<typename T1>
inline
bool
op_princomp::direct_princomp
  (
         Mat<typename T1::elem_type>&     coeff_out,
  const Base<typename T1::elem_type, T1>& X
  )
  {
  arma_debug_sigprint();
  
  typedef typename T1::elem_type eT;
  typedef typename T1::pod_type   T;
  
  const unwrap<T1>    Y( X.get_ref() );
  const Mat<eT>& in = Y.M;
  
  if(in.n_elem != 0)
    {
    Mat<eT> tmp = in; tmp.each_row() -= mean(in);
    
    // singular value decomposition
    Mat<eT> U;
    Col< T> s;
    
    const bool svd_ok = (in.n_rows >= in.n_cols) ? svd_econ(U, s, coeff_out, tmp) : svd(U, s, coeff_out, tmp);
    
    if(svd_ok == false)  { return false; }
    }
  else
    {
    coeff_out.eye(in.n_cols, in.n_cols);
    }
  
  return true;
  }



template<typename T1>
inline
void
op_princomp::apply
  (
        Mat<typename T1::elem_type>& out,
  const Op<T1,op_princomp>&          in
  )
  {
  arma_debug_sigprint();
  
  const bool status = op_princomp::direct_princomp(out, in.m);
  
  if(status == false)
    {
    out.soft_reset();
    
    arma_stop_runtime_error("princomp(): decomposition failed");
    }
  }



//! @}