// SPDX-License-Identifier: Apache-2.0 // // Copyright 2008-2016 Conrad Sanderson (http://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 // http://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_sp_mean //! @{ template<typename T1> inline void op_sp_mean::apply(Mat<typename T1::elem_type>& out, const mtSpReduceOp<typename T1::elem_type, T1, op_sp_mean>& in) { arma_debug_sigprint(); const uword dim = in.aux_uword_a; arma_conform_check( (dim > 1), "mean(): parameter 'dim' must be 0 or 1" ); const SpProxy<T1> p(in.m); const uword p_n_rows = p.get_n_rows(); const uword p_n_cols = p.get_n_cols(); if( (p_n_rows == 0) || (p_n_cols == 0) || (p.get_n_nonzero() == 0) ) { if(dim == 0) { out.zeros((p_n_rows > 0) ? 1 : 0, p_n_cols); } if(dim == 1) { out.zeros(p_n_rows, (p_n_cols > 0) ? 1 : 0); } return; } op_sp_mean::apply_fast(out, p, dim); } template<typename T1> inline void op_sp_mean::apply_fast ( Mat<typename T1::elem_type>& out, const SpProxy<T1>& p, const uword dim ) { arma_debug_sigprint(); typedef typename T1::elem_type eT; typedef typename T1::pod_type T; const uword p_n_rows = p.get_n_rows(); const uword p_n_cols = p.get_n_cols(); if(dim == 0) // find the mean in each column { arma_debug_print("op_sp_mean::apply_fast(): dim = 0"); out.zeros(1, p_n_cols); eT* out_mem = out.memptr(); if(SpProxy<T1>::use_iterator) { typename SpProxy<T1>::const_iterator_type it = p.begin(); const uword N = p.get_n_nonzero(); for(uword i=0; i < N; ++i) { out_mem[it.col()] += (*it); ++it; } out /= T(p_n_rows); } else { for(uword col = 0; col < p_n_cols; ++col) { out_mem[col] = arrayops::accumulate ( &p.get_values()[p.get_col_ptrs()[col]], p.get_col_ptrs()[col + 1] - p.get_col_ptrs()[col] ) / T(p_n_rows); } } } else if(dim == 1) // find the mean in each row { arma_debug_print("op_sp_mean::apply_fast(): dim = 1"); out.zeros(p_n_rows, 1); eT* out_mem = out.memptr(); typename SpProxy<T1>::const_iterator_type it = p.begin(); const uword N = p.get_n_nonzero(); for(uword i=0; i < N; ++i) { out_mem[it.row()] += (*it); ++it; } out /= T(p_n_cols); } if(out.internal_has_nonfinite()) { op_sp_mean::apply_slow(out, p, dim); } } template<typename T1> inline void op_sp_mean::apply_slow ( Mat<typename T1::elem_type>& out, const SpProxy<T1>& p, const uword dim ) { arma_debug_sigprint(); typedef typename T1::elem_type eT; const uword p_n_rows = p.get_n_rows(); const uword p_n_cols = p.get_n_cols(); if(dim == 0) // find the mean in each column { arma_debug_print("op_sp_mean::apply_slow(): dim = 0"); out.zeros(1, p_n_cols); for(uword col = 0; col < p_n_cols; ++col) { // Do we have to use an iterator or can we use memory directly? if(SpProxy<T1>::use_iterator) { typename SpProxy<T1>::const_iterator_type it = p.begin_col(col); typename SpProxy<T1>::const_iterator_type end = p.begin_col(col + 1); const uword n_zero = p_n_rows - (end.pos() - it.pos()); out.at(0,col) = op_sp_mean::iterator_mean(it, end, n_zero, eT(0)); } else { out.at(0,col) = op_sp_mean::direct_mean ( &p.get_values()[p.get_col_ptrs()[col]], p.get_col_ptrs()[col + 1] - p.get_col_ptrs()[col], p_n_rows ); } } } else if(dim == 1) // find the mean in each row { arma_debug_print("op_sp_mean::apply_slow(): dim = 1"); out.zeros(p_n_rows, 1); for(uword row = 0; row < p_n_rows; ++row) { // We must use an iterator regardless of how it is stored. typename SpProxy<T1>::const_row_iterator_type it = p.begin_row(row); typename SpProxy<T1>::const_row_iterator_type end = p.end_row(row); const uword n_zero = p_n_cols - (end.pos() - it.pos()); out.at(row,0) = op_sp_mean::iterator_mean(it, end, n_zero, eT(0)); } } } template<typename eT> inline eT op_sp_mean::direct_mean ( const eT* const X, const uword length, const uword N ) { arma_debug_sigprint(); typedef typename get_pod_type<eT>::result T; const eT result = ((length > 0) && (N > 0)) ? eT(arrayops::accumulate(X, length) / T(N)) : eT(0); return arma_isfinite(result) ? result : op_sp_mean::direct_mean_robust(X, length, N); } template<typename eT> inline eT op_sp_mean::direct_mean_robust ( const eT* const X, const uword length, const uword N ) { arma_debug_sigprint(); typedef typename get_pod_type<eT>::result T; uword i, j; eT r_mean = eT(0); const uword diff = (N - length); // number of zeros for(i = 0, j = 1; j < length; i += 2, j += 2) { const eT Xi = X[i]; const eT Xj = X[j]; r_mean += (Xi - r_mean) / T(diff + j); r_mean += (Xj - r_mean) / T(diff + j + 1); } if(i < length) { const eT Xi = X[i]; r_mean += (Xi - r_mean) / T(diff + i + 1); } return r_mean; } template<typename T1> inline typename T1::elem_type op_sp_mean::mean_all(const SpBase<typename T1::elem_type, T1>& X) { arma_debug_sigprint(); typedef typename T1::elem_type eT; SpProxy<T1> p(X.get_ref()); if(p.get_n_elem() == 0) { arma_conform_check(true, "mean(): object has no elements"); return Datum<eT>::nan; } if(SpProxy<T1>::use_iterator) { typename SpProxy<T1>::const_iterator_type it = p.begin(); typename SpProxy<T1>::const_iterator_type end = p.end(); return op_sp_mean::iterator_mean(it, end, p.get_n_elem() - p.get_n_nonzero(), typename T1::elem_type(0)); } else // use_iterator == false; that is, we can directly access the values array { return op_sp_mean::direct_mean(p.get_values(), p.get_n_nonzero(), p.get_n_elem()); } } template<typename T1, typename spop_type> inline typename T1::elem_type op_sp_mean::mean_all(const SpOp<T1, spop_type>& expr) { arma_debug_sigprint(); typedef typename T1::elem_type eT; constexpr bool is_vectorise = \ (is_same_type<spop_type, spop_vectorise_row>::yes) || (is_same_type<spop_type, spop_vectorise_col>::yes) || (is_same_type<spop_type, spop_vectorise_all>::yes); if(is_vectorise) { return op_sp_mean::mean_all(expr.m); } const SpMat<eT> tmp = expr; return op_sp_mean::mean_all(tmp); } template<typename T1, typename eT> inline eT op_sp_mean::iterator_mean(T1& it, const T1& end, const uword n_zero, const eT junk) { arma_debug_sigprint(); arma_ignore(junk); typedef typename get_pod_type<eT>::result T; eT acc = eT(0); T1 backup_it(it); // in case we have to use robust iterator_mean const uword it_begin_pos = it.pos(); while(it != end) { acc += (*it); ++it; } const uword count = n_zero + (it.pos() - it_begin_pos); const eT result = (count > 0) ? eT(acc / T(count)) : eT(0); return arma_isfinite(result) ? result : op_sp_mean::iterator_mean_robust(backup_it, end, n_zero, eT(0)); } template<typename T1, typename eT> inline eT op_sp_mean::iterator_mean_robust(T1& it, const T1& end, const uword n_zero, const eT junk) { arma_debug_sigprint(); arma_ignore(junk); typedef typename get_pod_type<eT>::result T; eT r_mean = eT(0); const uword it_begin_pos = it.pos(); while(it != end) { r_mean += ((*it - r_mean) / T(n_zero + (it.pos() - it_begin_pos) + 1)); ++it; } return r_mean; } //! @}