/**
 * @file krls.h
 * @brief krls
 * @author Cat (null.null.null@qq.com)
 * @version 0.1
 * @date 2021-09-17
 * 
 * Copyright: Baosight Co. Ltd.
 * DO NOT COPY/USE WITHOUT PERMISSION
 * 
 */
#pragma once
#include <eqpalg/regression/train_result.h>
#include "mix_cc/algorithm/split.h"
#include <vector>
#include <memory>
namespace regression {
using namespace dlib;
/**
 * @brief krls 回归预测
 * @tparam rows 
 * @tparam cols 
 */
template <size_t rows = 0, size_t cols = 0>
class krls_t {
 public:
  using SampleType = matrix<double, rows, cols>;
  using KernelType = radial_basis_kernel<SampleType>;
  using SamplesType = std::vector<SampleType>;
  using LabelsType = std::vector<double>;
  using TrainedResult = TrainResult<krls<KernelType>, rows, cols>;

 private:
  double split_precent_;

 public:
  krls_t() { this->split_precent_ = 0.7; }

  explicit krls_t(double split_percent) {
    this->split_precent_ = split_percent;
  }

  /**
   * @brief 学习样本，产生判断函数
   * @param  gama             My Param doc
   * @param  tr_samples       My Param doc
   * @param  tr_labels        My Param doc
   * @return TrainedResult 
   */
  TrainedResult learn(const double gama, const SamplesType& tr_samples,
                      const LabelsType& tr_labels) {
    krls<KernelType> trainer(KernelType(0.1), 0.001);
    size_t split_size = tr_samples.size() * (split_precent_);
    auto splitted_sample = mix_cc::split(tr_samples, {split_size});
    auto splitted_label = mix_cc::split(tr_labels, {split_size});
    auto train_size = splitted_sample[0].size();
    for (size_t i = 0; i < train_size; i++) {
      trainer.train(splitted_sample[0][i], splitted_label[0][i]);
    }
    TrainedResult result;
    result.set_predict_func(std::make_unique<krls<KernelType>>(trainer));
    std::vector<double> loo_values;
    auto valide_size = splitted_sample[1].size();
    for (auto i = 0; i < valide_size; i++) {
      loo_values.push_back(trainer(splitted_sample[1][i]));
    }
    result.err_rate = mean_squared_error(splitted_label[1], loo_values);
    return result;
  }
};
}  // namespace regression