Program Listing for File solver.hpp

Return to documentation for file (/home/docs/checkouts/readthedocs.org/user_builds/qulacs-rtd/checkouts/latest/src/vqcsim/solver.hpp)

#pragma once

#include <Eigen/Dense>
#include <cppsim/circuit_builder.hpp>
#include <cppsim/simulator.hpp>
#include <cppsim/type.hpp>
#include <cppsim/utility.hpp>
#include <functional>
#include <vector>

#include "differential.hpp"
#include "optimizer.hpp"
#include "problem.hpp"

class QuantumCircuitEnergyMinimizationSolver {
private:
    ParametricQuantumCircuit* _circuit;
    const std::function<ParametricQuantumCircuit*(UINT, UINT)>*
        _circuit_construction;
    UINT _param_count;
    std::vector<double> _parameter;
    double loss;

public:
    bool verbose;
    QuantumCircuitEnergyMinimizationSolver(
        const std::function<ParametricQuantumCircuit*(UINT, UINT)>*
            circuit_generator,
        UINT param_count = 0) {
        _circuit_construction = circuit_generator;
        _param_count = param_count;
        _circuit = NULL;
        verbose = false;
    };
    virtual ~QuantumCircuitEnergyMinimizationSolver() {
        if (_circuit != NULL) {
            delete _circuit;
        }
    }

    virtual void solve(EnergyMinimizationProblem* instance,
        UINT max_iteration = 100, std::string optimizer_name = "GD",
        std::string differentiation_method = "HalfPi") {
        if (_circuit != NULL) {
            delete _circuit;
            _circuit = NULL;
        }
        _circuit =
            (*_circuit_construction)(instance->get_qubit_count(), _param_count);
        auto _simulator = new ParametricQuantumCircuitSimulator(_circuit);
        _param_count = _simulator->get_parametric_gate_count();

        _parameter = std::vector<double>(_param_count, 0.);
        std::vector<double> gradient(_param_count);
        Random random;
        random.set_seed(0);
        for (auto& val : _parameter) val = random.uniform() * acos(0.0) * 4;

        GradientBasedOptimizer* optimizer;
        QuantumCircuitGradientDifferentiation* differentiation;

        if (optimizer_name == "Adam") {
            optimizer = new AdamOptimizer(_param_count);
        } else if (optimizer_name == "GD") {
            optimizer = new GradientDecentOptimizer(_param_count);
        } else
            return;
        if (differentiation_method == "HalfPi") {
            differentiation = new GradientByHalfPi();
        }
        std::vector<double> old_param;
        for (UINT iteration = 0; iteration < max_iteration; ++iteration) {
            loss = differentiation->compute_gradient(
                _simulator, instance, _parameter, &gradient);

            if (verbose) {
                std::cout << " *** epoch " << iteration << " *** " << std::endl;
                std::cout << " * loss = " << loss << std::endl;
                old_param = _parameter;
            }

            optimizer->apply_gradient(&_parameter, gradient);

            if (verbose) {
                for (UINT i = 0; i < _param_count; ++i) {
                    std::cout << " ** id " << i << " para = " << old_param[i]
                              << " -> " << _parameter[i]
                              << " , grad = " << gradient[i] << std::endl;
                }
            }
        }
        delete optimizer;
        delete differentiation;
        delete _simulator;
    }
    virtual double get_loss() { return loss; }
    virtual std::vector<double> get_parameter() { return _parameter; }
    ParametricQuantumCircuitSimulator* get_quantum_circuit_simulator() {
        return new ParametricQuantumCircuitSimulator(_circuit);
    }
};

class DiagonalizationEnergyMinimizationSolver {
private:
    ParametricQuantumCircuit* _circuit;
    const std::function<ParametricQuantumCircuit*(UINT, UINT)>*
        _circuit_construction;
    UINT _param_count;
    std::vector<double> _parameter;
    double loss;

public:
    bool verbose;
    DiagonalizationEnergyMinimizationSolver() { verbose = false; };
    virtual ~DiagonalizationEnergyMinimizationSolver() {}

    virtual void solve(EnergyMinimizationProblem* instance) {
        const UINT qubit_count = instance->get_qubit_count();
        const UINT term_count = instance->get_term_count();
        const ITYPE matrix_dim = 1ULL << qubit_count;

        ComplexMatrix observable_matrix =
            ComplexMatrix::Zero(matrix_dim, matrix_dim);
        for (UINT term_index = 0; term_index < term_count; ++term_index) {
            auto Pauli_operator = instance->get_Pauli_operator(term_index);
            CPPCTYPE coef = Pauli_operator->get_coef();
            auto target_index_list = Pauli_operator->get_index_list();
            auto pauli_id_list = Pauli_operator->get_pauli_id_list();

            std::vector<UINT> whole_pauli_id_list(qubit_count, 0);
            for (UINT i = 0; i < target_index_list.size(); ++i) {
                whole_pauli_id_list[target_index_list[i]] = pauli_id_list[i];
            }

            ComplexMatrix pauli_matrix;
            get_Pauli_matrix(pauli_matrix, whole_pauli_id_list);
            observable_matrix += coef * pauli_matrix;
        }

        observable_matrix.eigenvalues();
        Eigen::SelfAdjointEigenSolver<ComplexMatrix> eigen_solver(
            observable_matrix);
        loss = eigen_solver.eigenvalues()[0];

        if (verbose)
            std::cout << "Eigenvalues : " << std::endl
                      << eigen_solver.eigenvalues() << std::endl;
    }
    virtual double get_loss() { return loss; }
};