full_gen_tikhonov.h

Go to the documentation of this file.

/*
 * Copyright 2021 MusicScience37 (Kenta Kabashima)
 *
 * 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.
 */
#pragma once
 
#include <type_traits>
 
#include <Eigen/Core>
#include <Eigen/Householder>
 
#include "num_collect/base/concepts/dense_matrix.h"
#include "num_collect/base/exception.h"
#include "num_collect/base/index_type.h"
#include "num_collect/base/precondition.h"
#include "num_collect/logging/log_tag_view.h"
#include "num_collect/logging/logging_macros.h"
#include "num_collect/regularization/explicit_regularized_solver_base.h"
#include "num_collect/regularization/tikhonov.h"
#include "num_collect/util/assert.h"
 
namespace num_collect::regularization {
 
constexpr auto full_gen_tikhonov_tag =
    logging::log_tag_view("num_collect::regularization::full_gen_tikhonov");
 
template <base::concepts::dense_matrix Coeff, base::concepts::dense_matrix Data>
class full_gen_tikhonov
    : public explicit_regularized_solver_base<full_gen_tikhonov<Coeff, Data>,
          Data> {
public:
    using base_type =
        explicit_regularized_solver_base<full_gen_tikhonov<Coeff, Data>, Data>;
 
    using typename base_type::data_type;
    using typename base_type::scalar_type;
 
    using coeff_type = Coeff;
 
    static_assert(std::is_same_v<typename coeff_type::Scalar,
        typename data_type::Scalar>);
    static_assert(data_type::RowsAtCompileTime == Eigen::Dynamic);
 
    full_gen_tikhonov() : base_type(full_gen_tikhonov_tag) {}
 
    void compute(const coeff_type& coeff, const data_type& data,
        const coeff_type& reg_coeff) {
        NUM_COLLECT_PRECONDITION(coeff.rows() == data.rows(), this->logger(),
            "The number of rows in the coefficient matrix must match the "
            "number of rows in data.");
        NUM_COLLECT_PRECONDITION(coeff.cols() == reg_coeff.cols(),
            this->logger(),
            "The number of columns in the coefficient matrix must match the "
            "number of columns in the coefficient matrix of the regularization "
            "term.");
        NUM_COLLECT_PRECONDITION(reg_coeff.rows() < reg_coeff.cols(),
            this->logger(),
            "Coefficient matrix for the regularization term must have rows "
            "less than columns.");
 
        // How can I implement those complex formulas with good variable names.
 
        const index_type m = coeff.rows();
        const index_type n = coeff.cols();
        const index_type p = reg_coeff.rows();
 
        Eigen::ColPivHouseholderQR<coeff_type> qr_reg_adj;
        qr_reg_adj.compute(reg_coeff.adjoint());
        NUM_COLLECT_PRECONDITION(qr_reg_adj.rank() >= qr_reg_adj.cols(),
            this->logger(), "reg_coeff must have full row rank.");
        const coeff_type v = qr_reg_adj.householderQ();
 
        Eigen::ColPivHouseholderQR<coeff_type> qr_coeff_v2;
        qr_coeff_v2.compute(coeff * v.rightCols(n - p));
        NUM_COLLECT_PRECONDITION(qr_coeff_v2.rank() >= qr_coeff_v2.cols(),
            this->logger(),
            "reg_coeff and coeff must not have common elements "
            "other than zero in their kernel.");
        const coeff_type q = qr_coeff_v2.householderQ();
 
        const coeff_type coeff_arr =
            qr_reg_adj.solve(coeff.adjoint() * q.rightCols(m - n + p))
                .adjoint();
        const data_type data_arr = q.rightCols(m - n + p).adjoint() * data;
        tikhonov_.compute(coeff_arr, data_arr);
 
        const coeff_type coeff_v2_inv_coeff = qr_coeff_v2.solve(coeff);
        const coeff_type i_minus_v2_coeff_v2_inv_coeff =
            coeff_type::Identity(n, n) -
            v.rightCols(n - p) * coeff_v2_inv_coeff;
        coeff_actual_solution_ =
            qr_reg_adj.solve(i_minus_v2_coeff_v2_inv_coeff.adjoint()).adjoint();
 
        const data_type coeff_v2_inv_data = qr_coeff_v2.solve(data);
        offset_actual_solution_ = v.rightCols(n - p) * coeff_v2_inv_data;
    }
 
    void solve(const scalar_type& param, data_type& solution) const {
        data_type tikhonov_solution;
        tikhonov_.solve(param, tikhonov_solution);
        solution = coeff_actual_solution_ * tikhonov_solution +
            offset_actual_solution_;
    }
 
    [[nodiscard]] auto singular_values() const -> const
        typename Eigen::BDCSVD<coeff_type>::SingularValuesType& {
        return tikhonov_.singular_values();
    }
 
    [[nodiscard]] auto residual_norm(const scalar_type& param) const
        -> scalar_type {
        return tikhonov_.residual_norm(param);
    }
 
    [[nodiscard]] auto regularization_term(const scalar_type& param) const
        -> scalar_type {
        return tikhonov_.regularization_term(param);
    }
 
    [[nodiscard]] auto first_derivative_of_residual_norm(
        const scalar_type& param) const -> scalar_type {
        return tikhonov_.first_derivative_of_residual_norm(param);
    }
 
    [[nodiscard]] auto first_derivative_of_regularization_term(
        const scalar_type& param) const -> scalar_type {
        return tikhonov_.first_derivative_of_regularization_term(param);
    }
 
    [[nodiscard]] auto second_derivative_of_residual_norm(
        const scalar_type& param) const -> scalar_type {
        return tikhonov_.second_derivative_of_residual_norm(param);
    }
 
    [[nodiscard]] auto second_derivative_of_regularization_term(
        const scalar_type& param) const -> scalar_type {
        return tikhonov_.second_derivative_of_regularization_term(param);
    }
 
    [[nodiscard]] auto sum_of_filter_factor(const scalar_type& param) const
        -> scalar_type {
        return tikhonov_.sum_of_filter_factor(param);
    }
 
    [[nodiscard]] auto data_size() const -> index_type {
        return tikhonov_.data_size();
    }
 
    [[nodiscard]] auto param_search_region() const
        -> std::pair<scalar_type, scalar_type> {
        return tikhonov_.param_search_region();
    }
 
    [[nodiscard]] auto internal_solver() const
        -> const tikhonov<coeff_type, data_type>& {
        return tikhonov_;
    }
 
private:
    tikhonov<coeff_type, data_type> tikhonov_{};
 
    Coeff coeff_actual_solution_{};
 
    Data offset_actual_solution_{};
};
 
}  // namespace num_collect::regularization