pypharm 1.5.2__py3-none-any.whl → 1.6.1__py3-none-any.whl

PyPharm/models/compartment_models.py

@@ -1,639 +1,985 @@
-from multiprocessing import shared_memory
-
-import numpy as np
-from scipy.integrate import solve_ivp, RK45
-from scipy.integrate import simps
-from scipy.optimize import minimize
-from PyPharm.algorithms.country_optimization import CountriesAlgorithm
-from PyPharm.algorithms.country_optimization_v2 import CountriesAlgorithm_v2
-from PyPharm.algorithms.genetic_optimization import GeneticAlgorithm
-from numba import njit
-import matplotlib.pyplot as plt
-
-
-class BaseCompartmentModel:
-
-    configuration_matrix_target = None
-    outputs_target = None
-    volumes_target = None
-    _optim = False
-    numba_option = False
-
-    def __init__(self, configuration_matrix, outputs, volumes=None, numba_option=False, use_shared_memory=False):
-        """
-        Базовая камерная модель для описания фармакокинетики системы
-
-        Неизвестные параметры при необходимости задаются как None
-        например configuration_matrix = [[0, 1], [None, 0]]
-
-        Args:
-            configuration_matrix: Настроечная матрица модели, отображающая константы перехода между матрицами
-            outputs: Вектор констант перехода во вне камер
-            volumes: Объемы камер
-        """
-        self.configuration_matrix = np.array(configuration_matrix)
-        self.configuration_matrix_target_count = 0
-        if np.any(self.configuration_matrix == None):
-            self.configuration_matrix_target = np.where(self.configuration_matrix == None)
-            self.configuration_matrix_target_count = np.sum(self.configuration_matrix == None)
-        self.outputs = np.array(outputs)
-        self.outputs_target_count = 0
-        if np.any(self.outputs == None):
-            self.outputs_target = np.where(self.outputs == None)
-            self.outputs_target_count = np.sum(self.outputs == None)
-        if not volumes:
-            self.volumes = np.ones(self.outputs.size)
-        else:
-            self.volumes = np.array(volumes)
-        self.volumes_target_count = 0
-        if np.any(self.volumes == None):
-            self.volumes_target = np.where(self.volumes == None)
-            self.volumes_target_count = np.sum(self.volumes == None)
-        self.last_result = None
-        self.numba_option = numba_option
-        self.use_shared_memory = use_shared_memory
-        if self.use_shared_memory:
-            self.memory_size = 2 + self.configuration_matrix_target_count + self.outputs_target_count + self.volumes_target_count
-            self.memory = shared_memory.ShareableList(self.memory_size * [None])
-            self.memory_name = self.memory.shm.name
-
-    def __del__(self):
-        if getattr(self, 'memory', None):
-            self.memory.shm.close()
-            self.memory.shm.unlink()
-
-    def _compartment_model(self, t, c):
-        """
-        Функция для расчета камерной модели
-
-        Args:
-            t: Текущее время
-            c: Вектор концентраций
-
-        Returns:
-            Вектор изменений концентраций (c) в момент времени (t)
-        """
-        dc_dt = (self.configuration_matrix.T @ (c * self.volumes) \
-            - self.configuration_matrix.sum(axis=1) * (c * self.volumes)\
-            - self.outputs * (c * self.volumes)) / self.volumes
-        return dc_dt
-
-    @staticmethod
-    @njit
-    def _numba_compartment_model(t, c, configuration_matrix, outputs, volumes):
-        """
-        Функция для расчета камерной модели
-
-        Args:
-            t: Текущее время
-            c: Вектор концентраций
-
-        Returns:
-            Вектор изменений концентраций (c) в момент времени (t)
-        """
-        dc_dt = (configuration_matrix.T @ (c * volumes) \
-            - configuration_matrix.sum(axis=1) * (c * volumes)\
-            - outputs * (c * volumes)) / volumes
-        return dc_dt
-
-    def __call__(self, t_max, c0=None, d=None, compartment_number=None, max_step=0.01, t_eval=None):
-        """
-        Расчет кривых концентраций по фармакокинетической модели
-
-        Args:
-            t_max: Предельное время расчета
-            c0: Вектор нулевых концентраций
-            d: Вводимая доза
-            compartment_number: Номер камеры в которую вводится доза
-            max_step: Максимальный шаг при решении СДУ
-            t_eval: Временные точки, в которых необходимо молучить решение
-
-        Returns:
-            Результат работы решателя scipy solve_ivp
-        """
-        if not self._optim:
-            assert (not any([self.configuration_matrix_target, self.outputs_target, self.volumes_target])), \
-                "It is impossible to make a calculation with unknown parameters"
-        assert any([c0 is not None, d, compartment_number is not None]), "Need to set c0 or d and compartment_number"
-        if c0 is None:
-            assert all([d, compartment_number is not None]), "Need to set d and compartment_number"
-            c0 = np.zeros(self.outputs.size)
-            c0[compartment_number] = d / self.volumes[compartment_number]
-        else:
-            c0 = np.array(c0)
-        ts = [0, t_max]
-        self.last_result = solve_ivp(
-            fun=self._compartment_model if not self.numba_option else lambda t, c: self._numba_compartment_model(t, c, self.configuration_matrix.astype(np.float64), self.outputs.astype(np.float64), self.volumes.astype(np.float64)),
-            t_span=ts,
-            y0=c0,
-            max_step=max_step,
-            t_eval=t_eval,
-            method='LSODA'
-        )
-        return self.last_result
-
-    def get_kinetic_params(self, t_max, d, compartment_number, max_step=0.01):
-        one_hour_result = self(t_max=1, d=d, compartment_number=compartment_number, max_step=max_step)
-        auc_1h = simps(one_hour_result.y[compartment_number], one_hour_result.t)
-        self(t_max=t_max, d=d, compartment_number=compartment_number, max_step=max_step)
-        auc = simps(self.last_result.y[compartment_number], self.last_result.t)
-        result_dict = {
-            'c_max': self.last_result.y[compartment_number].max(),
-            'V': self.volumes[compartment_number] if self.volumes is not None else None,
-            'AUC': auc,
-            'AUC_1h': auc_1h,
-            'Cl': d / auc
-        }
-        return result_dict
-
-
-    def load_data_from_list(self, x):
-        if self.configuration_matrix_target:
-            self.configuration_matrix[self.configuration_matrix_target] = x[:self.configuration_matrix_target_count]
-        if self.outputs_target:
-            self.outputs[self.outputs_target] = x[
-                                                self.configuration_matrix_target_count:self.configuration_matrix_target_count + self.outputs_target_count]
-        if self.volumes_target:
-            self.volumes[self.volumes_target] = x[self.configuration_matrix_target_count + self.outputs_target_count:self.configuration_matrix_target_count + self.outputs_target_count + self.volumes_target_count]
-
-    def _target_function(self, x, max_step=0.01, metric='R2'):
-        """
-        Функция расчета значения целевой функции
-
-        Args:
-            x: Значение искомых параметров модели
-            max_step: Максимальный шаг при решении СДУ
-
-        Returns:
-            Значение целевой функции, характеризующее отклонение от эксперементальных данных
-        """
-        self.load_data_from_list(x)
-        c0 = self.c0
-        if c0 is None:
-            c0 = np.zeros(self.outputs.size)
-            c0[self.compartment_number] = self.d / self.volumes[self.compartment_number]
-        self(
-            t_max=np.max(self.teoretic_x),
-            c0=c0,
-            t_eval=self.teoretic_x,
-            max_step=max_step
-        )
-        target_results = self.last_result.y[tuple(self.know_compartments), :]
-        if metric == 'R2':
-            return np.sum(np.sum(self.w * ((target_results - self.teoretic_y) ** 2), axis=1) / np.sum((self.teoretic_avg - self.teoretic_y) ** 2, axis=1))
-        elif metric == 'norm':
-            return np.linalg.norm(target_results - self.teoretic_y)
-        else:
-            return np.sum(np.sum(self.w * ((target_results - self.teoretic_y) ** 2), axis=1) / np.sum((self.teoretic_avg - self.teoretic_y) ** 2, axis=1))
-
-
-    def load_optimization_data(self, teoretic_x, teoretic_y, know_compartments, w = None, c0=None, d=None, compartment_number=None):
-        """
-        Функция загрузки в модель эксперементальных данных
-
-        Args:
-            teoretic_x: Вектор временных точек теоретических значений
-            teoretic_y: Матрица с теоретическими значениями
-            know_compartments: Вектор с номерами камер, по которым есть данные
-            c0: Вектор нулевых концентраций
-            d: Вводимая доза
-            compartment_number: Номер камеры в которую вводится доза
-
-        Returns:
-            None
-        """
-        self.teoretic_x = np.array(teoretic_x)
-        self.teoretic_y = np.array(teoretic_y)
-        self.know_compartments = know_compartments
-        self.teoretic_avg = np.average(self.teoretic_y, axis=1)
-        self.teoretic_avg = np.repeat(self.teoretic_avg, self.teoretic_x.size)
-        self.teoretic_avg = np.reshape(self.teoretic_avg, self.teoretic_y.shape)
-        assert any([c0, d, compartment_number is not None]), "Need to set c0 or d and compartment_number"
-        if not c0:
-            assert all([d, compartment_number is not None]), "Need to set d and compartment_number"
-            self.d = d
-            self.compartment_number = compartment_number
-            self.c0 = None
-        else:
-            self.c0 = np.array(c0)
-        self.w = np.ones(self.teoretic_y.shape) if w is None else np.array(w)
-
-    def optimize(self, method=None, user_method=None, method_is_func=True,
-                 optimization_func_name='__call__', max_step=0.01, metric='R2', **kwargs):
-        """
-        Функция оптимизации модели
-
-        Args:
-            method: Метод оптимизации, любой доступный minimize + 'country_optimization' и 'country_optimization_v2'
-            max_step: Максимальный шаг при решении СДУ
-            **kwargs: Дополнительные именованные аргументы
-
-        Returns:
-            None
-        """
-        self._optim = True
-        f = lambda x: self._target_function(x, max_step=max_step, metric=metric)
-        if user_method is not None:
-            if method_is_func:
-                x = user_method(f, **kwargs)
-            else:
-                optimization_obj = user_method(f, **kwargs)
-                x = getattr(optimization_obj, optimization_func_name)()
-        else:
-            if method == 'country_optimization':
-                CA = CountriesAlgorithm(
-                    f=f,
-                    memory_list=getattr(self, 'memory', None),
-                    **kwargs
-                )
-                CA.start()
-                x = CA.countries[0].population[0].x
-            elif method == 'country_optimization_v2':
-                CA = CountriesAlgorithm_v2(
-                    f=f,
-                    **kwargs
-                )
-                CA.start()
-                x = CA.countries[0].population[0].x
-            elif method == 'GA':
-                CA = GeneticAlgorithm(
-                    f=f,
-                    **kwargs
-                )
-                x = CA.start()
-            else:
-                res = minimize(
-                    fun=f,
-                    method=method,
-                    **kwargs
-                )
-                x = res.x
-        if self.configuration_matrix_target:
-            self.configuration_matrix[self.configuration_matrix_target] = x[:self.configuration_matrix_target_count]
-        if self.outputs_target:
-            self.outputs[self.outputs_target] = x[
-                                                self.configuration_matrix_target_count:self.configuration_matrix_target_count + self.outputs_target_count]
-        if self.volumes_target:
-            self.volumes[self.volumes_target] = x[self.configuration_matrix_target_count + self.outputs_target_count:self.configuration_matrix_target_count + self.outputs_target_count + self.volumes_target_count]
-        self.configuration_matrix_target = None
-        self.outputs_target = None
-        self.volumes_target = None
-        self._optim = False
-        return x
-
-    def plot_model(self, compartment_numbers=None, compartment_names={}, left=None, right=None, y_lims={}, **kwargs):
-        """
-        Функция для построения графиков модели
-
-        Args:
-            compartment_numbers: Камеры, которые нужно отобразить (если не указать, отобразим все)
-            compartment_names: Имена камер
-        """
-        if compartment_numbers:
-            compartment_numbers = np.array(compartment_numbers)
-        else:
-            compartment_numbers = np.arange(self.outputs.size)
-        self(**kwargs)
-        for i in compartment_numbers:
-            if hasattr(self, "teoretic_x") and hasattr(self, "teoretic_y") and i in self.know_compartments:
-                plt.plot(self.teoretic_x, self.teoretic_y[self.know_compartments.index(i)], "*r")
-            plt.plot(self.last_result.t, self.last_result.y[i])
-            plt.title(compartment_names.get(i, i))
-            plt.xlim(left=left, right=right)
-            if y_lims.get(i):
-                plt.ylim(y_lims.get(i))
-            plt.grid()
-            plt.show()
-
-
-class MagicCompartmentModel(BaseCompartmentModel):
-
-    need_magic_optimization = False
-
-    def __init__(self, configuration_matrix, outputs, volumes=None, magic_coefficient=1, exclude_compartments=[], numba_option=False, use_shared_memory=False):
-        super().__init__(configuration_matrix, outputs, volumes, numba_option, use_shared_memory)
-        self.magic_coefficient = magic_coefficient
-        self.exclude_compartments = np.array(exclude_compartments)
-        self.need_magic_optimization = self.magic_coefficient is None
-        if getattr(self, "memory", None):
-            self.memory.shm.close()
-            self.memory.shm.unlink()
-            self.memory_size += int(self.need_magic_optimization)
-            self.memory = shared_memory.ShareableList(
-                sequence=self.memory_size * [None]
-            )
-            self.memory_name = self.memory.shm.name
-
-    def __call__(self, t_max, c0=None, d=None, compartment_number=None, max_step=0.01, t_eval=None):
-        if not self._optim and not self.magic_coefficient:
-            raise Exception("Magic_coefficient parameter not specified")
-        res = super().__call__(t_max, c0, d, compartment_number, max_step, t_eval)
-        magic_arr = np.ones(self.configuration_matrix.shape[0]) * self.magic_coefficient
-        if self.exclude_compartments:
-            magic_arr[self.exclude_compartments] = 1
-        magic_arr = np.repeat(magic_arr, res.y.shape[1])
-        magic_arr = np.reshape(magic_arr, res.y.shape)
-        res.y = magic_arr * res.y
-        self.last_result = res
-        return res
-
-    def load_data_from_list(self, x):
-        super().load_data_from_list(x)
-        if self.need_magic_optimization:
-            self.magic_coefficient = x[-1]
-        
-    def optimize(self, method=None, user_method=None, method_is_func=True,
-                 optimization_func_name='__call__', max_step=0.01, **kwargs):
-        x = super().optimize(method, user_method, method_is_func, optimization_func_name, max_step, **kwargs)
-        if self.need_magic_optimization:
-            self.magic_coefficient = x[-1]
-            self.need_magic_optimization = False
-        return x
-
-
-class ReleaseCompartmentModel(BaseCompartmentModel):
-
-    need_v_release_optimization = False
-    release_parameters_target = None
-    accumulation_parameters_target = None
-
-    class ReleaseRK45(RK45):
-
-        def __init__(self, fun, t0, y0, t_bound, release_function, compartment_number, c0, max_step=np.inf,
-                     with_accumulate=False, accumulation_function=None, accumulation_type=1, rtol=1e-3, atol=1e-6, vectorized=False,
-                     first_step=None,  **extraneous):
-            super().__init__(fun, t0, y0, t_bound, max_step=max_step,
-                     rtol=rtol, atol=atol, vectorized=vectorized,
-                     first_step=first_step, **extraneous)
-            self.release_function = release_function
-            self.compartment_number = compartment_number
-            self.c0 = c0
-            self.old_release_correction = 0
-            self.old_accumulation_correction = c0
-            self.with_accumulate = with_accumulate
-            self.accumulation_function = accumulation_function
-            self.accumulation_type = accumulation_type
-
-        def _step_impl(self):
-            result = super()._step_impl()
-            release_correction = self.release_function(self.t, self.c0)
-            minus_accumulation = 0
-            if self.with_accumulate:
-                if self.accumulation_type == 1:
-                    accumulation_correction = self.accumulation_function(self.t, self.c0 )
-                    minus_accumulation = self.old_accumulation_correction - accumulation_correction
-                elif self.accumulation_type == 2:
-                    coef_accumulation = self.accumulation_function(self.t, self.c0) / self.c0
-            plus_release = release_correction - self.old_release_correction
-            all_corrections = plus_release
-            if self.accumulation_type == 1:
-                if plus_release - minus_accumulation > 0:
-                    all_corrections = plus_release - minus_accumulation
-            elif self.accumulation_type == 2:
-                all_corrections *= coef_accumulation
-            self.y[self.compartment_number] += all_corrections
-            self.old_release_correction = release_correction
-            if self.with_accumulate and self.accumulation_type == 1:
-                self.old_accumulation_correction = accumulation_correction
-            return result
-
-    def __init__(self, v_release, release_parameters, release_compartment,
-                 release_function=None, with_accumulate=False, accumulation_function=None,
-                 accumulation_parameters=[None], accumulation_type=1, *args, **kwargs):
-        """
-        Камерная модель с высвобождением для описания фармакокинетики системы
-
-        Неизвестные параметры при необходимости задаются как None
-        например configuration_matrix = [[0, 1], [None, 0]]
-
-        Args:
-            configuration_matrix: Настроечная матрица модели, отображающая константы перехода между матрицами
-            outputs: Вектор констант перехода во вне камер
-            volumes: Объемы камер
-            v_release: Объем гепотетической камеры из которой происходит высвобождение
-            release_parameters: Параметры функции высвобождения
-            release_compartment: Номер камеры в которую происходит высвобождение
-            release_function: Функция высвобождения по умолчанию f(t,m,b,c) = c0 * c * t ** b / (t ** b + m)
-        """
-        super().__init__(*args, **kwargs)
-        self.release_parameters = np.array(release_parameters)
-        self.release_parameters_target_count = 0
-        if np.any(self.release_parameters == None):
-            self.release_parameters_target = np.where(self.release_parameters == None)
-            self.release_parameters_target_count = np.sum(self.release_parameters == None)
-        self.v_release = v_release
-        if self.v_release is None:
-            self.need_v_release_optimization = True
-        self.release_compartment = release_compartment
-        self.release_function = release_function
-        self.with_accumulate = with_accumulate
-        self.accumulation_type = accumulation_type
-
-        if self.with_accumulate:
-            self.accumulation_function = accumulation_function
-            self.accumulation_parameters = np.array(accumulation_parameters)
-            if np.any(self.accumulation_parameters == None):
-                self.accumulation_parameters_target = np.where(self.accumulation_parameters == None)
-                self.accumulation_parameters_target_count = np.sum(self.accumulation_parameters == None)
-
-        if getattr(self, "memory", None):
-            self.memory.shm.close()
-            self.memory.shm.unlink()
-            self.memory_size += self.release_parameters_target_count + int(self.need_v_release_optimization)
-            self.memory = shared_memory.ShareableList(
-                sequence=self.memory_size * [None]
-            )
-            self.memory_name = self.memory.shm.name
-
-    def load_data_from_list(self, x):
-        super().load_data_from_list(x)
-        s = self.configuration_matrix_target_count + self.outputs_target_count + self.volumes_target_count
-        if self.release_parameters_target:
-            self.release_parameters[self.release_parameters_target] = x[s:s + self.release_parameters_target_count]
-        if self.need_v_release_optimization:
-            self.v_release = x[s + self.release_parameters_target_count]
-        if self.with_accumulate:
-            if self.accumulation_parameters_target:
-                s += self.release_parameters_target_count + int(self.need_v_release_optimization)
-                self.accumulation_parameters[self.accumulation_parameters_target] = x[s:s + self.accumulation_parameters_target_count]
-
-    def _default_release_function(self, t, c0):
-        """
-        Функция для поправки на высвобождение
-        """
-        m, b, c = self.release_parameters
-        return c0 * c * t ** b / (t ** b + m)
-
-    def get_release_function(self):
-        if self.release_function is not None:
-            return lambda t, c0: self.release_function(t, c0, *self.release_parameters)
-        else:
-            return self._default_release_function
-
-    def _default_accumulation_function(self, t, c0):
-        """
-        Функция для поправки на накопление
-        """
-        k, = self.accumulation_parameters
-        return c0 * np.exp(-k * t)
-
-    def get_accumulation_function(self):
-        if self.accumulation_function is not None:
-            return lambda t, c0: self.accumulation_function(t, c0, *self.accumulation_parameters)
-        else:
-            return self._default_accumulation_function
-
-    def __call__(self, t_max, c0=None, d=None, max_step=0.01, t_eval=None, **kwargs):
-        """
-                Расчет кривых концентраций по фармакокинетической модели
-
-                Args:
-                    t_max: Предельное время расчета
-                    c0: Начальная концентрация в камере из которой высвобождается вещество
-                    d: Вводимая доза
-                    max_step: Максимальный шаг при решении СДУ
-                    t_eval: Временные точки, в которых необходимо молучить решение
-
-                Returns:
-                    Результат работы решателя scipy solve_ivp
-                """
-        if not self._optim:
-            assert (not any([self.configuration_matrix_target, self.outputs_target, self.volumes_target])), \
-                "It is impossible to make a calculation with unknown parameters"
-        assert any([c0 is not None, d]), "Need to set c0 or d and compartment_number"
-        if c0 is None:
-            assert d, "Need to set d"
-            c0 = d / self.v_release
-        ts = [0, t_max]
-        y0 = np.zeros(self.outputs.shape)
-        self.last_result = solve_ivp(
-            fun=self._compartment_model if
-            not self.numba_option
-            else lambda t, c: self._numba_compartment_model(t, c,
-                                                             self.configuration_matrix.astype(
-                                                                 np.float64),
-                                                             self.outputs.astype(
-                                                                 np.float64),
-                                                             self.volumes.astype(
-                                                     np.float64)),
-            t_span=ts,
-            y0=y0,
-            max_step=max_step,
-            t_eval=t_eval,
-            method=self.ReleaseRK45,
-            release_function=self.get_release_function(),
-            compartment_number=self.release_compartment,
-            with_accumulate=self.with_accumulate,
-            accumulation_function=self.get_accumulation_function() if self.with_accumulate else None,
-            accumulation_type=self.accumulation_type,
-            c0=c0
-        )
-        self.last_result.model_realized = c0 - self.get_release_function()(self.last_result.t, c0)
-        if self.with_accumulate:
-            model_accumulation = self.get_accumulation_function()(self.last_result.t, c0)
-            if self.accumulation_type == 1:
-                self.last_result.model_realized = model_accumulation - self.get_release_function()(self.last_result.t, c0)
-            elif self.accumulation_type == 2:
-                accumulation_coeffs = model_accumulation / c0
-                self.last_result.model_realized= accumulation_coeffs * self.get_release_function()(self.last_result.t, c0)
-                self.last_result.model_realized = model_accumulation - self.last_result.model_realized
-        return self.last_result
-
-    def _target_function(self, x, max_step=0.01, metric='R2'):
-        """
-        Функция расчета значения целевой функции
-
-        Args:
-            x: Значение искомых параметров модели
-            max_step: Максимальный шаг при решении СДУ
-
-        Returns:
-            Значение целевой функции, характеризующее отклонение от эксперементальных данных
-        """
-        self.load_data_from_list(x)
-        c0 = self.c0
-        if c0 is None:
-            c0 = self.d / self.v_release
-        self(
-            t_max=np.max(self.teoretic_x),
-            c0=c0,
-            t_eval=self.teoretic_x,
-            max_step=max_step
-        )
-        target_results = self.last_result.y[tuple(self.know_compartments), :]
-        if metric == 'R2':
-            plus = 0
-            if self.teoretic_realized is not None:
-                model_realized = self.last_result.model_realized
-                plus = np.sum(((model_realized - self.teoretic_realized) ** 2) / ((self.teoretic_realized - self.teoretic_realized_avg) ** 2))
-            return plus + np.sum(np.sum(self.w * ((target_results - self.teoretic_y) ** 2), axis=1) / np.sum((self.teoretic_avg - self.teoretic_y) ** 2, axis=1))
-        elif metric == 'norm':
-            plus = 0
-            if self.teoretic_realized is not None:
-                model_realized = self.last_result.model_realized
-                plus = np.linalg.norm(self.teoretic_realized - model_realized)
-            return plus + np.linalg.norm(target_results - self.teoretic_y)
-        else:
-            return np.sum(np.sum(self.w * ((target_results - self.teoretic_y) ** 2), axis=1) / np.sum((self.teoretic_avg - self.teoretic_y) ** 2, axis=1))
-
-    def load_optimization_data(self, teoretic_x, teoretic_y, know_compartments,
-                               w = None, c0=None, d=None, compartment_number=None, teoretic_realized=None):
-        """
-        Функция загрузки в модель эксперементальных данных
-
-        Args:
-            teoretic_x: Вектор временных точек теоретических значений
-            teoretic_y: Матрица с теоретическими значениями
-            know_compartments: Вектор с номерами камер, по которым есть данные
-            c0: Начальная концентрация в камере из которой высвобождается вещество
-            d: Вводимая доза
-
-        Returns:
-            None
-        """
-        self.teoretic_x = np.array(teoretic_x)
-        self.teoretic_y = np.array(teoretic_y)
-        self.teoretic_realized = np.array(teoretic_realized) if teoretic_realized is not None else teoretic_realized
-        if teoretic_realized is not None:
-            self.teoretic_realized_avg = np.average(self.teoretic_realized)
-        self.know_compartments = know_compartments
-        self.teoretic_avg = np.average(self.teoretic_y, axis=1)
-        self.teoretic_avg = np.repeat(self.teoretic_avg, self.teoretic_x.size)
-        self.teoretic_avg = np.reshape(self.teoretic_avg, self.teoretic_y.shape)
-        assert any([c0, d, compartment_number is not None]), "Need to set c0 or d and compartment_number"
-        if not c0:
-            assert all([d, compartment_number is not None]), "Need to set d and compartment_number"
-            self.d = d
-            self.c0 = None
-        else:
-            self.c0 = np.array(c0)
-        self.w = np.ones(self.teoretic_y.shape) if w is None else np.array(w)
-
-    def optimize(self, method=None, user_method=None, method_is_func=True,
-                 optimization_func_name='__call__', max_step=0.01, **kwargs):
-        x = super().optimize(method, user_method, method_is_func, optimization_func_name, max_step, **kwargs)
-        s = self.configuration_matrix_target_count + self.outputs_target_count + self.volumes_target_count
-        if self.release_parameters_target:
-            self.release_parameters[self.release_parameters_target] = x[s:s + self.release_parameters_target_count]
-        if self.need_v_release_optimization:
-            self.v_release = x[s:s + self.release_parameters_target_count + 1]
-            self.need_v_release_optimization = False
-        if self.with_accumulate:
-            if self.accumulation_parameters_target:
-                s += self.release_parameters_target_count + int(self.need_v_release_optimization)
-                self.accumulation_parameters[self.accumulation_parameters_target] = x[s:s + self.accumulation_parameters_target_count]
-        return x
-
-    def plot_model(self, compartment_numbers=None, compartment_names={},
-                   left=None, right=None, y_lims={}, plot_accumulation=False, **kwargs):
-        super().plot_model(compartment_numbers, compartment_names, left, right, y_lims, **kwargs)
-        if plot_accumulation:
-            if hasattr(self, "teoretic_x") and hasattr(self, "teoretic_realized"):
-                plt.plot(self.teoretic_x, self.teoretic_realized, "*r")
-            plt.plot(self.last_result.t, self.last_result.model_realized)
-            plt.title(compartment_names.get('realized', 'realized'))
-            plt.xlim(left=left, right=right)
-            if y_lims.get('realized'):
-                plt.ylim(y_lims.get('realized'))
-            plt.grid()
-            plt.show()
+from multiprocessing import shared_memory
+
+import numpy as np
+from scipy.integrate import solve_ivp, RK45
+from scipy.integrate import simps
+from scipy.optimize import minimize
+from PyPharm.algorithms.country_optimization import CountriesAlgorithm
+from PyPharm.algorithms.country_optimization_v2 import CountriesAlgorithm_v2
+from PyPharm.algorithms.genetic_optimization import GeneticAlgorithm
+from numba import njit
+import matplotlib.pyplot as plt
+
+BIG_VALUE = 2 ** 24
+
+
+class BaseCompartmentModel:
+
+    configuration_matrix_target = None
+    outputs_target = None
+    volumes_target = None
+    _optim = False
+    numba_option = False
+
+    def __init__(self, configuration_matrix, outputs, volumes=None, numba_option=False, use_shared_memory=False):
+        """
+        Базовая камерная модель для описания фармакокинетики системы
+
+        Неизвестные параметры при необходимости задаются как None
+        например configuration_matrix = [[0, 1], [None, 0]]
+
+        Args:
+            configuration_matrix: Настроечная матрица модели, отображающая константы перехода между матрицами
+            outputs: Вектор констант перехода во вне камер
+            volumes: Объемы камер
+        """
+        self.configuration_matrix = np.array(configuration_matrix)
+        self.configuration_matrix_target_count = 0
+        if np.any(self.configuration_matrix == None):
+            self.configuration_matrix_target = np.where(self.configuration_matrix == None)
+            self.configuration_matrix_target_count = np.sum(self.configuration_matrix == None)
+        self.outputs = np.array(outputs)
+        self.outputs_target_count = 0
+        if np.any(self.outputs == None):
+            self.outputs_target = np.where(self.outputs == None)
+            self.outputs_target_count = np.sum(self.outputs == None)
+        if not volumes:
+            self.volumes = np.ones(self.outputs.size)
+        else:
+            self.volumes = np.array(volumes)
+        self.volumes_target_count = 0
+        if np.any(self.volumes == None):
+            self.volumes_target = np.where(self.volumes == None)
+            self.volumes_target_count = np.sum(self.volumes == None)
+        self.last_result = None
+        self.numba_option = numba_option
+        self.use_shared_memory = use_shared_memory
+        if self.use_shared_memory:
+            self.memory_size = 2 + self.configuration_matrix_target_count + self.outputs_target_count + self.volumes_target_count
+            self.memory = shared_memory.ShareableList(self.memory_size * [None])
+            self.memory_name = self.memory.shm.name
+
+    def __del__(self):
+        if getattr(self, 'memory', None):
+            self.memory.shm.close()
+            self.memory.shm.unlink()
+
+    def _compartment_model(self, t, c):
+        """
+        Функция для расчета камерной модели
+
+        Args:
+            t: Текущее время
+            c: Вектор концентраций
+
+        Returns:
+            Вектор изменений концентраций (c) в момент времени (t)
+        """
+        dc_dt = (self.configuration_matrix.T @ (c * self.volumes) \
+            - self.configuration_matrix.sum(axis=1) * (c * self.volumes)\
+            - self.outputs * (c * self.volumes)) / self.volumes
+        return dc_dt
+
+    @staticmethod
+    @njit
+    def _numba_compartment_model(t, c, configuration_matrix, outputs, volumes):
+        """
+        Функция для расчета камерной модели
+
+        Args:
+            t: Текущее время
+            c: Вектор концентраций
+
+        Returns:
+            Вектор изменений концентраций (c) в момент времени (t)
+        """
+        dc_dt = (configuration_matrix.T @ (c * volumes) \
+            - configuration_matrix.sum(axis=1) * (c * volumes)\
+            - outputs * (c * volumes)) / volumes
+        return dc_dt
+
+    def __call__(self, t_max, c0=None, d=None, compartment_number=None, max_step=0.01, t_eval=None):
+        """
+        Расчет кривых концентраций по фармакокинетической модели
+
+        Args:
+            t_max: Предельное время расчета
+            c0: Вектор нулевых концентраций
+            d: Вводимая доза
+            compartment_number: Номер камеры в которую вводится доза
+            max_step: Максимальный шаг при решении СДУ
+            t_eval: Временные точки, в которых необходимо молучить решение
+
+        Returns:
+            Результат работы решателя scipy solve_ivp
+        """
+        if not self._optim:
+            assert (not any([self.configuration_matrix_target, self.outputs_target, self.volumes_target])), \
+                "It is impossible to make a calculation with unknown parameters"
+        assert any([c0 is not None, d, compartment_number is not None]), "Need to set c0 or d and compartment_number"
+        if c0 is None:
+            assert all([d, compartment_number is not None]), "Need to set d and compartment_number"
+            c0 = np.zeros(self.outputs.size)
+            c0[compartment_number] = d / self.volumes[compartment_number]
+        else:
+            c0 = np.array(c0)
+        ts = [0, t_max]
+        self.last_result = solve_ivp(
+            fun=self._compartment_model if not self.numba_option else lambda t, c: self._numba_compartment_model(t, c, self.configuration_matrix.astype(np.float64), self.outputs.astype(np.float64), self.volumes.astype(np.float64)),
+            t_span=ts,
+            y0=c0,
+            max_step=max_step,
+            t_eval=t_eval,
+            method='LSODA'
+        )
+        return self.last_result
+
+    def get_kinetic_params(self, t_max, d, compartment_number, max_step=0.01):
+        one_hour_result = self(t_max=1, d=d, compartment_number=compartment_number, max_step=max_step)
+        auc_1h = simps(one_hour_result.y[compartment_number], one_hour_result.t)
+        self(t_max=t_max, d=d, compartment_number=compartment_number, max_step=max_step)
+        auc = simps(self.last_result.y[compartment_number], self.last_result.t)
+        result_dict = {
+            'c_max': self.last_result.y[compartment_number].max(),
+            'V': self.volumes[compartment_number] if self.volumes is not None else None,
+            'AUC': auc,
+            'AUC_1h': auc_1h,
+            'Cl': d / auc
+        }
+        return result_dict
+
+
+    def load_data_from_list(self, x):
+        if self.configuration_matrix_target:
+            self.configuration_matrix[self.configuration_matrix_target] = x[:self.configuration_matrix_target_count]
+        if self.outputs_target:
+            self.outputs[self.outputs_target] = x[
+                                                self.configuration_matrix_target_count:self.configuration_matrix_target_count + self.outputs_target_count]
+        if self.volumes_target:
+            self.volumes[self.volumes_target] = x[self.configuration_matrix_target_count + self.outputs_target_count:self.configuration_matrix_target_count + self.outputs_target_count + self.volumes_target_count]
+
+    def _target_function(self, x, max_step=0.01, metric='R2'):
+        """
+        Функция расчета значения целевой функции
+
+        Args:
+            x: Значение искомых параметров модели
+            max_step: Максимальный шаг при решении СДУ
+
+        Returns:
+            Значение целевой функции, характеризующее отклонение от эксперементальных данных
+        """
+        self.load_data_from_list(x)
+        c0 = self.c0
+        if c0 is None:
+            c0 = np.zeros(self.outputs.size)
+            c0[self.compartment_number] = self.d / self.volumes[self.compartment_number]
+        self(
+            t_max=np.max(self.teoretic_x),
+            c0=c0,
+            t_eval=self.teoretic_x,
+            max_step=max_step
+        )
+        target_results = self.last_result.y[tuple(self.know_compartments), :]
+        if metric == 'R2':
+            return np.sum(np.sum(self.w * ((target_results - self.teoretic_y) ** 2), axis=1) / np.sum((self.teoretic_avg - self.teoretic_y) ** 2, axis=1))
+        elif metric == 'norm':
+            return np.linalg.norm(target_results - self.teoretic_y)
+        else:
+            return np.sum(np.sum(self.w * ((target_results - self.teoretic_y) ** 2), axis=1) / np.sum((self.teoretic_avg - self.teoretic_y) ** 2, axis=1))
+
+
+    def load_optimization_data(self, teoretic_x, teoretic_y, know_compartments, w = None, c0=None, d=None, compartment_number=None):
+        """
+        Функция загрузки в модель эксперементальных данных
+
+        Args:
+            teoretic_x: Вектор временных точек теоретических значений
+            teoretic_y: Матрица с теоретическими значениями
+            know_compartments: Вектор с номерами камер, по которым есть данные
+            c0: Вектор нулевых концентраций
+            d: Вводимая доза
+            compartment_number: Номер камеры в которую вводится доза
+
+        Returns:
+            None
+        """
+        self.teoretic_x = np.array(teoretic_x)
+        self.teoretic_y = np.array(teoretic_y)
+        self.know_compartments = know_compartments
+        self.teoretic_avg = np.average(self.teoretic_y, axis=1)
+        self.teoretic_avg = np.repeat(self.teoretic_avg, self.teoretic_x.size)
+        self.teoretic_avg = np.reshape(self.teoretic_avg, self.teoretic_y.shape)
+        assert any([c0, d, compartment_number is not None]), "Need to set c0 or d and compartment_number"
+        if not c0:
+            assert all([d, compartment_number is not None]), "Need to set d and compartment_number"
+            self.d = d
+            self.compartment_number = compartment_number
+            self.c0 = None
+        else:
+            self.c0 = np.array(c0)
+        self.w = np.ones(self.teoretic_y.shape) if w is None else np.array(w)
+
+    def optimize(self, method=None, user_method=None, method_is_func=True,
+                 optimization_func_name='__call__', max_step=0.01, metric='R2', **kwargs):
+        """
+        Функция оптимизации модели
+
+        Args:
+            method: Метод оптимизации, любой доступный minimize + 'country_optimization' и 'country_optimization_v2'
+            max_step: Максимальный шаг при решении СДУ
+            **kwargs: Дополнительные именованные аргументы
+
+        Returns:
+            None
+        """
+        self._optim = True
+        f = lambda x: self._target_function(x, max_step=max_step, metric=metric)
+        if user_method is not None:
+            if method_is_func:
+                x = user_method(f, **kwargs)
+            else:
+                optimization_obj = user_method(f, **kwargs)
+                x = getattr(optimization_obj, optimization_func_name)()
+        else:
+            if method == 'country_optimization':
+                CA = CountriesAlgorithm(
+                    f=f,
+                    memory_list=getattr(self, 'memory', None),
+                    **kwargs
+                )
+                CA.start()
+                x = CA.countries[0].population[0].x
+            elif method == 'country_optimization_v2':
+                CA = CountriesAlgorithm_v2(
+                    f=f,
+                    **kwargs
+                )
+                CA.start()
+                x = CA.countries[0].population[0].x
+            elif method == 'GA':
+                CA = GeneticAlgorithm(
+                    f=f,
+                    **kwargs
+                )
+                x = CA.start()
+            else:
+                res = minimize(
+                    fun=f,
+                    method=method,
+                    **kwargs
+                )
+                x = res.x
+        if self.configuration_matrix_target:
+            self.configuration_matrix[self.configuration_matrix_target] = x[:self.configuration_matrix_target_count]
+        if self.outputs_target:
+            self.outputs[self.outputs_target] = x[
+                                                self.configuration_matrix_target_count:self.configuration_matrix_target_count + self.outputs_target_count]
+        if self.volumes_target:
+            self.volumes[self.volumes_target] = x[self.configuration_matrix_target_count + self.outputs_target_count:self.configuration_matrix_target_count + self.outputs_target_count + self.volumes_target_count]
+        self.configuration_matrix_target = None
+        self.outputs_target = None
+        self.volumes_target = None
+        self._optim = False
+        return x
+
+    def plot_model(self, compartment_numbers=None, compartment_names=None, left=None, right=None, y_lims={}, **kwargs):
+        """
+        Функция для построения графиков модели
+
+        Args:
+            compartment_numbers: Камеры, которые нужно отобразить (если не указать, отобразим все)
+            compartment_names: Имена камер
+        """
+        if compartment_numbers:
+            compartment_numbers = np.array(compartment_numbers)
+        else:
+            compartment_numbers = np.arange(self.outputs.size)
+        if not compartment_names:
+            compartment_names = {}
+        if not y_lims:
+            y_lims = {}
+        self(**kwargs)
+        for i in compartment_numbers:
+            if hasattr(self, "teoretic_x") and hasattr(self, "teoretic_y") and i in self.know_compartments:
+                plt.plot(self.teoretic_x, self.teoretic_y[self.know_compartments.index(i)], "*r")
+            plt.plot(self.last_result.t, self.last_result.y[i])
+            plt.title(compartment_names.get(i, i))
+            plt.xlim(left=left, right=right)
+            if y_lims.get(i):
+                plt.ylim(y_lims.get(i))
+            plt.grid()
+            try:
+                plt.show()
+            except AttributeError:
+                plt.savefig(f'{compartment_names.get(i, str(i))}.png')
+                plt.cla()
+
+
+class MagicCompartmentModel(BaseCompartmentModel):
+
+    need_magic_optimization = False
+
+    def __init__(self, configuration_matrix, outputs, volumes=None, magic_coefficient=1, exclude_compartments=None, numba_option=False, use_shared_memory=False):
+        super().__init__(configuration_matrix, outputs, volumes, numba_option, use_shared_memory)
+        self.magic_coefficient = magic_coefficient
+        self.exclude_compartments = np.array(exclude_compartments) if bool(exclude_compartments) else np.array([])
+        self.need_magic_optimization = self.magic_coefficient is None
+        if getattr(self, "memory", None):
+            self.memory.shm.close()
+            self.memory.shm.unlink()
+            self.memory_size += int(self.need_magic_optimization)
+            self.memory = shared_memory.ShareableList(
+                sequence=self.memory_size * [None]
+            )
+            self.memory_name = self.memory.shm.name
+
+    def __call__(self, t_max, c0=None, d=None, compartment_number=None, max_step=0.01, t_eval=None):
+        if not self._optim and not self.magic_coefficient:
+            raise Exception("Magic_coefficient parameter not specified")
+        res = super().__call__(t_max, c0, d, compartment_number, max_step, t_eval)
+        magic_arr = np.ones(self.configuration_matrix.shape[0]) * self.magic_coefficient
+        if self.exclude_compartments:
+            magic_arr[self.exclude_compartments] = 1
+        magic_arr = np.repeat(magic_arr, res.y.shape[1])
+        magic_arr = np.reshape(magic_arr, res.y.shape)
+        res.y = magic_arr * res.y
+        self.last_result = res
+        return res
+
+    def load_data_from_list(self, x):
+        super().load_data_from_list(x)
+        if self.need_magic_optimization:
+            self.magic_coefficient = x[-1]
+        
+    def optimize(self, method=None, user_method=None, method_is_func=True,
+                 optimization_func_name='__call__', max_step=0.01, **kwargs):
+        x = super().optimize(method, user_method, method_is_func, optimization_func_name, max_step, **kwargs)
+        if self.need_magic_optimization:
+            self.magic_coefficient = x[-1]
+            self.need_magic_optimization = False
+        return x
+
+
+class ReleaseCompartmentModel(BaseCompartmentModel):
+
+    need_v_release_optimization = False
+    release_parameters_target = None
+    accumulation_parameters_target = None
+
+    class ReleaseRK45(RK45):
+
+        def __init__(self, fun, t0, y0, t_bound, release_function, compartment_number, c0, max_step=np.inf,
+                     with_accumulate=False, accumulation_function=None, accumulation_type=1, rtol=1e-3, atol=1e-6, vectorized=False,
+                     first_step=None,  **extraneous):
+            super().__init__(fun, t0, y0, t_bound, max_step=max_step,
+                     rtol=rtol, atol=atol, vectorized=vectorized,
+                     first_step=first_step, **extraneous)
+            self.release_function = release_function
+            self.compartment_number = compartment_number
+            self.c0 = c0
+            self.old_release_correction = 0
+            self.old_accumulation_correction = c0
+            self.with_accumulate = with_accumulate
+            self.accumulation_function = accumulation_function
+            self.accumulation_type = accumulation_type
+
+        def _step_impl(self):
+            result = super()._step_impl()
+            release_correction = self.release_function(self.t, self.c0)
+            minus_accumulation = 0
+            if self.with_accumulate:
+                if self.accumulation_type == 1:
+                    accumulation_correction = self.accumulation_function(self.t, self.c0 )
+                    minus_accumulation = self.old_accumulation_correction - accumulation_correction
+                elif self.accumulation_type == 2:
+                    coef_accumulation = self.accumulation_function(self.t, self.c0) / self.c0
+            plus_release = release_correction - self.old_release_correction
+            all_corrections = plus_release
+            if self.accumulation_type == 1:
+                if plus_release - minus_accumulation > 0:
+                    all_corrections = plus_release - minus_accumulation
+            elif self.accumulation_type == 2:
+                all_corrections *= coef_accumulation
+            self.y[self.compartment_number] += all_corrections
+            self.old_release_correction = release_correction
+            if self.with_accumulate and self.accumulation_type == 1:
+                self.old_accumulation_correction = accumulation_correction
+            return result
+
+    def __init__(self, v_release, release_parameters, release_compartment,
+                 release_function=None, with_accumulate=False, accumulation_function=None,
+                 accumulation_parameters=[None], accumulation_type=1, *args, **kwargs):
+        """
+        Камерная модель с высвобождением для описания фармакокинетики системы
+
+        Неизвестные параметры при необходимости задаются как None
+        например configuration_matrix = [[0, 1], [None, 0]]
+
+        Args:
+            configuration_matrix: Настроечная матрица модели, отображающая константы перехода между матрицами
+            outputs: Вектор констант перехода во вне камер
+            volumes: Объемы камер
+            v_release: Объем гепотетической камеры из которой происходит высвобождение
+            release_parameters: Параметры функции высвобождения
+            release_compartment: Номер камеры в которую происходит высвобождение
+            release_function: Функция высвобождения по умолчанию f(t,m,b,c) = c0 * c * t ** b / (t ** b + m)
+        """
+        super().__init__(*args, **kwargs)
+        self.release_parameters = np.array(release_parameters)
+        self.release_parameters_target_count = 0
+        if np.any(self.release_parameters == None):
+            self.release_parameters_target = np.where(self.release_parameters == None)
+            self.release_parameters_target_count = np.sum(self.release_parameters == None)
+        self.v_release = v_release
+        if self.v_release is None:
+            self.need_v_release_optimization = True
+        self.release_compartment = release_compartment
+        self.release_function = release_function
+        self.with_accumulate = with_accumulate
+        self.accumulation_type = accumulation_type
+
+        if self.with_accumulate:
+            self.accumulation_function = accumulation_function
+            self.accumulation_parameters = np.array(accumulation_parameters)
+            if np.any(self.accumulation_parameters == None):
+                self.accumulation_parameters_target = np.where(self.accumulation_parameters == None)
+                self.accumulation_parameters_target_count = np.sum(self.accumulation_parameters == None)
+
+        if getattr(self, "memory", None):
+            self.memory.shm.close()
+            self.memory.shm.unlink()
+            self.memory_size += self.release_parameters_target_count + int(self.need_v_release_optimization)
+            self.memory = shared_memory.ShareableList(
+                sequence=self.memory_size * [None]
+            )
+            self.memory_name = self.memory.shm.name
+
+    def load_data_from_list(self, x):
+        super().load_data_from_list(x)
+        s = self.configuration_matrix_target_count + self.outputs_target_count + self.volumes_target_count
+        if self.release_parameters_target:
+            self.release_parameters[self.release_parameters_target] = x[s:s + self.release_parameters_target_count]
+        if self.need_v_release_optimization:
+            self.v_release = x[s + self.release_parameters_target_count]
+        if self.with_accumulate:
+            if self.accumulation_parameters_target:
+                s += self.release_parameters_target_count + int(self.need_v_release_optimization)
+                self.accumulation_parameters[self.accumulation_parameters_target] = x[s:s + self.accumulation_parameters_target_count]
+
+    def _default_release_function(self, t, c0):
+        """
+        Функция для поправки на высвобождение
+        """
+        m, b, c = self.release_parameters
+        return c0 * c * t ** b / (t ** b + m)
+
+    def get_release_function(self):
+        if self.release_function is not None:
+            return lambda t, c0: self.release_function(t, c0, *self.release_parameters)
+        else:
+            return self._default_release_function
+
+    def _default_accumulation_function(self, t, c0):
+        """
+        Функция для поправки на накопление
+        """
+        k, = self.accumulation_parameters
+        return c0 * np.exp(-k * t)
+
+    def get_accumulation_function(self):
+        if self.accumulation_function is not None:
+            return lambda t, c0: self.accumulation_function(t, c0, *self.accumulation_parameters)
+        else:
+            return self._default_accumulation_function
+
+    def __call__(self, t_max, c0=None, d=None, max_step=0.01, t_eval=None, **kwargs):
+        """
+                Расчет кривых концентраций по фармакокинетической модели
+
+                Args:
+                    t_max: Предельное время расчета
+                    c0: Начальная концентрация в камере из которой высвобождается вещество
+                    d: Вводимая доза
+                    max_step: Максимальный шаг при решении СДУ
+                    t_eval: Временные точки, в которых необходимо молучить решение
+
+                Returns:
+                    Результат работы решателя scipy solve_ivp
+                """
+        if not self._optim:
+            assert (not any([self.configuration_matrix_target, self.outputs_target, self.volumes_target])), \
+                "It is impossible to make a calculation with unknown parameters"
+        assert any([c0 is not None, d]), "Need to set c0 or d and compartment_number"
+        if c0 is None:
+            assert d, "Need to set d"
+            c0 = d / self.v_release
+        ts = [0, t_max]
+        y0 = np.zeros(self.outputs.shape)
+        self.last_result = solve_ivp(
+            fun=self._compartment_model if
+            not self.numba_option
+            else lambda t, c: self._numba_compartment_model(t, c,
+                                                             self.configuration_matrix.astype(
+                                                                 np.float64),
+                                                             self.outputs.astype(
+                                                                 np.float64),
+                                                             self.volumes.astype(
+                                                     np.float64)),
+            t_span=ts,
+            y0=y0,
+            max_step=max_step,
+            t_eval=t_eval,
+            method=self.ReleaseRK45,
+            release_function=self.get_release_function(),
+            compartment_number=self.release_compartment,
+            with_accumulate=self.with_accumulate,
+            accumulation_function=self.get_accumulation_function() if self.with_accumulate else None,
+            accumulation_type=self.accumulation_type,
+            c0=c0
+        )
+        self.last_result.model_realized = c0 - self.get_release_function()(self.last_result.t, c0)
+        if self.with_accumulate:
+            model_accumulation = self.get_accumulation_function()(self.last_result.t, c0)
+            if self.accumulation_type == 1:
+                self.last_result.model_realized = model_accumulation - self.get_release_function()(self.last_result.t, c0)
+            elif self.accumulation_type == 2:
+                accumulation_coeffs = model_accumulation / c0
+                self.last_result.model_realized= accumulation_coeffs * self.get_release_function()(self.last_result.t, c0)
+                self.last_result.model_realized = model_accumulation - self.last_result.model_realized
+        return self.last_result
+
+    def _target_function(self, x, max_step=0.01, metric='R2'):
+        """
+        Функция расчета значения целевой функции
+
+        Args:
+            x: Значение искомых параметров модели
+            max_step: Максимальный шаг при решении СДУ
+
+        Returns:
+            Значение целевой функции, характеризующее отклонение от эксперементальных данных
+        """
+        self.load_data_from_list(x)
+        c0 = self.c0
+        if c0 is None:
+            c0 = self.d / self.v_release
+        self(
+            t_max=np.max(self.teoretic_x),
+            c0=c0,
+            t_eval=self.teoretic_x,
+            max_step=max_step
+        )
+        target_results = self.last_result.y[tuple(self.know_compartments), :]
+        if metric == 'R2':
+            plus = 0
+            if self.teoretic_realized is not None:
+                model_realized = self.last_result.model_realized
+                plus = np.sum(((model_realized - self.teoretic_realized) ** 2) / ((self.teoretic_realized - self.teoretic_realized_avg) ** 2))
+            return plus + np.sum(np.sum(self.w * ((target_results - self.teoretic_y) ** 2), axis=1) / np.sum((self.teoretic_avg - self.teoretic_y) ** 2, axis=1))
+        elif metric == 'norm':
+            plus = 0
+            if self.teoretic_realized is not None:
+                model_realized = self.last_result.model_realized
+                plus = np.linalg.norm(self.teoretic_realized - model_realized)
+            return plus + np.linalg.norm(target_results - self.teoretic_y)
+        else:
+            return np.sum(np.sum(self.w * ((target_results - self.teoretic_y) ** 2), axis=1) / np.sum((self.teoretic_avg - self.teoretic_y) ** 2, axis=1))
+
+    def load_optimization_data(self, teoretic_x, teoretic_y, know_compartments,
+                               w = None, c0=None, d=None, compartment_number=None, teoretic_realized=None):
+        """
+        Функция загрузки в модель эксперементальных данных
+
+        Args:
+            teoretic_x: Вектор временных точек теоретических значений
+            teoretic_y: Матрица с теоретическими значениями
+            know_compartments: Вектор с номерами камер, по которым есть данные
+            c0: Начальная концентрация в камере из которой высвобождается вещество
+            d: Вводимая доза
+
+        Returns:
+            None
+        """
+        self.teoretic_x = np.array(teoretic_x)
+        self.teoretic_y = np.array(teoretic_y)
+        self.teoretic_realized = np.array(teoretic_realized) if teoretic_realized is not None else teoretic_realized
+        if teoretic_realized is not None:
+            self.teoretic_realized_avg = np.average(self.teoretic_realized)
+        self.know_compartments = know_compartments
+        self.teoretic_avg = np.average(self.teoretic_y, axis=1)
+        self.teoretic_avg = np.repeat(self.teoretic_avg, self.teoretic_x.size)
+        self.teoretic_avg = np.reshape(self.teoretic_avg, self.teoretic_y.shape)
+        assert any([c0, d, compartment_number is not None]), "Need to set c0 or d and compartment_number"
+        if not c0:
+            assert all([d, compartment_number is not None]), "Need to set d and compartment_number"
+            self.d = d
+            self.c0 = None
+        else:
+            self.c0 = np.array(c0)
+        self.w = np.ones(self.teoretic_y.shape) if w is None else np.array(w)
+
+    def optimize(self, method=None, user_method=None, method_is_func=True,
+                 optimization_func_name='__call__', max_step=0.01, **kwargs):
+        x = super().optimize(method, user_method, method_is_func, optimization_func_name, max_step, **kwargs)
+        s = self.configuration_matrix_target_count + self.outputs_target_count + self.volumes_target_count
+        if self.release_parameters_target:
+            self.release_parameters[self.release_parameters_target] = x[s:s + self.release_parameters_target_count]
+        if self.need_v_release_optimization:
+            self.v_release = x[s:s + self.release_parameters_target_count + 1]
+            self.need_v_release_optimization = False
+        if self.with_accumulate:
+            if self.accumulation_parameters_target:
+                s += self.release_parameters_target_count + int(self.need_v_release_optimization)
+                self.accumulation_parameters[self.accumulation_parameters_target] = x[s:s + self.accumulation_parameters_target_count]
+        return x
+
+    def plot_model(self, compartment_numbers=None, compartment_names={},
+                   left=None, right=None, y_lims={}, plot_accumulation=False, **kwargs):
+        super().plot_model(compartment_numbers, compartment_names, left, right, y_lims, **kwargs)
+        if plot_accumulation:
+            if hasattr(self, "teoretic_x") and hasattr(self, "teoretic_realized"):
+                plt.plot(self.teoretic_x, self.teoretic_realized, "*r")
+            plt.plot(self.last_result.t, self.last_result.model_realized)
+            plt.title(compartment_names.get('realized', 'realized'))
+            plt.xlim(left=left, right=right)
+            if y_lims.get('realized'):
+                plt.ylim(y_lims.get('realized'))
+            plt.grid()
+            plt.show()
+
+
+class TwoSubstancesCompartmentModel(MagicCompartmentModel):
+
+    released_configuration_matrix_target = None
+    released_outputs_target = None
+    release_parameters_target = None
+    has_teoretic_y = False
+    has_teoretic_released = False
+
+    class TwoSubstancesRK45(RK45):
+
+        def __init__(self, fun, t0, y0, t_bound, release_function, max_step=np.inf,
+                     rtol=1e-3, atol=1e-6, vectorized=False,
+                     first_step=None,  **extraneous):
+            self.old_corrections = 0
+            super().__init__(fun, t0, y0, t_bound, max_step=max_step,
+                     rtol=rtol, atol=atol, vectorized=vectorized,
+                     first_step=first_step, **extraneous)
+            self.release_function = release_function
+
+        def _step_impl(self):
+            result = super()._step_impl()
+            release_correction: float = self.release_function(self.y, self.t)
+            correction = release_correction - self.old_corrections
+            c = self.y[:self.y.size // 2]
+            real_release_correction = np.append(-1 * correction * c, correction * c)
+            self.y += real_release_correction
+            self.old_corrections = release_correction
+            return result
+
+    def __init__(self, configuration_matrix, outputs, released_configuration_matrix, released_outputs,
+                 release_parameters, release_function=None,
+                 volumes=None, magic_coefficient=1, exclude_compartments=None,
+                 numba_option=False, use_shared_memory=False):
+        super().__init__(
+            configuration_matrix=configuration_matrix,
+            outputs=outputs,
+            volumes=volumes,
+            magic_coefficient=magic_coefficient,
+            exclude_compartments=exclude_compartments,
+            numba_option=numba_option,
+            use_shared_memory=use_shared_memory
+        )
+        self.released_configuration_matrix = np.array(released_configuration_matrix)
+        self.released_configuration_matrix_target_count = 0
+        if np.any(self.released_configuration_matrix == None):
+            self.released_configuration_matrix_target = np.where(self.released_configuration_matrix == None)
+            self.released_configuration_matrix_target_count = np.sum(self.released_configuration_matrix == None)
+        self.released_outputs = np.array(released_outputs)
+        self.released_outputs_target_count = 0
+        if np.any(self.released_outputs == None):
+            self.released_outputs_target = np.where(self.released_outputs == None)
+            self.released_outputs_target_count = np.sum(self.released_outputs == None)
+        self.release_parameters = np.array(release_parameters)
+        if np.any(self.release_parameters == None):
+            self.release_parameters_target = np.where(self.release_parameters == None)
+            self.release_parameters_target_count = np.sum(self.release_parameters == None)
+        self.release_function = release_function
+        if getattr(self, "memory", None):
+            self.memory.shm.close()
+            self.memory.shm.unlink()
+            self.memory_size += (self.release_parameters_target_count +
+                                 self.released_configuration_matrix_target_count + self.released_outputs_target_count)
+            self.memory = shared_memory.ShareableList(
+                sequence=self.memory_size * [None]
+            )
+            self.memory_name = self.memory.shm.name
+
+    def __call__(self, t_max, c0=None, d=None, compartment_number=None, max_step=0.01, t_eval=None, **kwargs):
+        """
+                Расчет кривых концентраций по фармакокинетической модели
+
+                Args:
+                    t_max: Предельное время расчета
+                    c0: Начальная концентрация в камере из которой высвобождается вещество
+                    d: Вводимая доза
+                    max_step: Максимальный шаг при решении СДУ
+                    t_eval: Временные точки, в которых необходимо молучить решение
+
+                Returns:
+                    Результат работы решателя scipy solve_ivp
+                """
+        if not self._optim:
+            assert (not any([self.configuration_matrix_target, self.outputs_target, self.volumes_target])), \
+                "It is impossible to make a calculation with unknown parameters"
+        assert any([c0 is not None, d]), "Need to set c0 or d and compartment_number"
+
+
+        if c0 is None:
+            assert all([d, compartment_number is not None]), "Need to set d and compartment_number"
+            c0 = np.zeros((2, self.outputs.size))
+            c0[0][compartment_number] = d / self.volumes[compartment_number]
+        else:
+            c0 = np.array(c0)
+        c0 = c0.reshape((1, c0.size))[0]
+        ts = [0, t_max]
+        res = solve_ivp(
+            fun=self._compartment_model if
+            not self.numba_option
+            else lambda t, c: self._numba_compartment_model(t, c,
+                                                             self.configuration_matrix.astype(
+                                                                 np.float64),
+                                                            self.released_configuration_matrix.astype(
+                                                                np.float64),
+                                                             self.outputs.astype(
+                                                                 np.float64),
+                                                            self.released_outputs.astype(
+                                                                np.float64),
+                                                             self.volumes.astype(
+                                                     np.float64)),
+            t_span=ts,
+            y0=c0,
+            max_step=max_step,
+            t_eval=t_eval,
+            method=self.TwoSubstancesRK45,
+            release_function=self.get_release_function()
+        )
+        magic_arr = np.ones(self.configuration_matrix.shape[0] + self.released_configuration_matrix.shape[0]) * self.magic_coefficient
+        if self.exclude_compartments.size:
+            magic_arr[self.exclude_compartments] = 1
+        magic_arr = np.repeat(magic_arr, res.y.shape[1])
+        magic_arr = np.reshape(magic_arr, res.y.shape)
+        res.y = magic_arr * res.y
+        self.last_result = res
+        return self.last_result
+
+    def _default_release_function(self, current_c, t):
+        """
+        Функция для поправки на высвобождение
+        """
+        # c_resh = c.reshape((2, c.size // 2))
+        # k = self.release_parameters[0]
+        # return k * c_resh[0]
+        m, b, c = self.release_parameters
+        return c * t ** b / (t ** b + m)
+
+    def get_release_function(self):
+        if self.release_function is not None:
+            return lambda c, t: self.release_function(c, t, *self.release_parameters)
+        else:
+            return self._default_release_function
+
+    def _compartment_model(self, t, c):
+        """
+        Функция для расчета камерной модели
+
+        Args:
+            t: Текущее время
+            c: Вектор концентраций
+
+        Returns:
+            Вектор изменений концентраций (c) в момент времени (t)
+        """
+        c_resh = c.reshape((2, c.size // 2))
+        c_primary = c_resh[0]
+        dc_dt = (self.configuration_matrix.T @ (c_primary * self.volumes) \
+            - self.configuration_matrix.sum(axis=1) * (c_primary * self.volumes)\
+            - self.outputs * (c_primary * self.volumes)) / self.volumes
+        c_released = c_resh[1]
+        dc_dt_released = (self.released_configuration_matrix.T @ (c_released * self.volumes) \
+                 - self.released_configuration_matrix.sum(axis=1) * (c_released * self.volumes) \
+                 - self.released_outputs * (c_released * self.volumes)) / self.volumes
+        np.append(dc_dt, dc_dt_released)
+
+    @staticmethod
+    @njit
+    def _numba_compartment_model(t, c, configuration_matrix, released_configuration_matrix, outputs, released_outputs, volumes):
+        """
+        Функция для расчета камерной модели
+
+        Args:
+            t: Текущее время
+            c: Вектор концентраций
+
+        Returns:
+            Вектор изменений концентраций (c) в момент времени (t)
+        """
+        c_resh = c.reshape((2, c.size // 2))
+        c_primary = c_resh[0]
+        dc_dt = (configuration_matrix.T @ (c_primary * volumes) \
+                 - configuration_matrix.sum(axis=1) * (c_primary * volumes) \
+                 - outputs * (c_primary * volumes)) / volumes
+        c_released = c_resh[1]
+        dc_dt_released = (released_configuration_matrix.T @ (c_released * volumes) \
+                          - released_configuration_matrix.sum(axis=1) * (c_released * volumes) \
+                          - released_outputs * (c_released * volumes)) / volumes
+        return np.append(dc_dt, dc_dt_released)
+
+    def load_optimization_data(self, teoretic_x, teoretic_y=None, teoretic_released=None, know_compartments=None, know_released_compartments=None,
+                               w=None, released_w=None, c0=None, d=None, compartment_number=None):
+        self.has_teoretic_y = bool(teoretic_y)
+        self.has_teoretic_released = bool(teoretic_released)
+        self.teoretic_x = np.array(teoretic_x)
+        if self.has_teoretic_y:
+            self.teoretic_y = np.array(teoretic_y if teoretic_y is not None else [[]])
+            self.know_compartments = know_compartments if know_compartments is not None else []
+            self.teoretic_avg = np.average(self.teoretic_y, axis=1)
+            self.teoretic_avg = np.repeat(self.teoretic_avg, self.teoretic_x.size)
+            self.teoretic_avg = np.reshape(self.teoretic_avg, self.teoretic_y.shape)
+            self.w = np.ones(self.teoretic_y.shape) if w is None else np.array(w)
+        if self.has_teoretic_released:
+            self.teoretic_released = np.array(teoretic_released if teoretic_released is not None else [[]])
+            self.know_released_compartments = know_released_compartments if know_released_compartments is not None else []
+            self.released_avg = np.average(self.teoretic_released, axis=1)
+            self.released_avg = np.repeat(self.released_avg, self.teoretic_x.size)
+            self.released_avg = np.reshape(self.released_avg, self.teoretic_released.shape)
+            self.released_w = np.ones(self.teoretic_released.shape) if released_w is None else np.array(released_w)
+
+        assert any([c0, d, compartment_number is not None]), "Need to set c0 or d and compartment_number"
+        if not c0:
+            assert all([d, compartment_number is not None]), "Need to set d and compartment_number"
+            self.d = d
+            self.compartment_number = compartment_number
+            self.c0 = None
+        else:
+            self.c0 = np.array(c0)
+
+
+    def load_data_from_list(self, x):
+        super().load_data_from_list(x)
+        n = self.configuration_matrix_target_count + self.outputs_target_count + self.volumes_target_count
+        if self.released_configuration_matrix_target:
+            self.released_configuration_matrix[self.released_configuration_matrix_target] = x[n:n + self.released_configuration_matrix_target_count]
+        n += self.released_configuration_matrix_target_count
+        if self.released_outputs_target:
+            self.released_outputs[self.released_outputs_target] = x[n:n + self.released_outputs_target_count]
+        n += self.released_outputs_target_count
+        if self.release_parameters_target:
+            self.release_parameters[self.release_parameters_target] = x[n:n + self.release_parameters_target_count]
+
+    def _target_function(self, x, max_step=0.01, metric='R2'):
+        self.load_data_from_list(x)
+        c0 = self.c0
+        if c0 is None:
+            c0 = np.zeros((2, self.outputs.size))
+            c0[0][self.compartment_number] = self.d / self.volumes[self.compartment_number]
+        self(
+            t_max=np.max(self.teoretic_x),
+            c0=c0,
+            t_eval=self.teoretic_x,
+            max_step=max_step
+        )
+        if not self.last_result.success:
+            return BIG_VALUE
+        if self.has_teoretic_y:
+            target_results = self.last_result.y[tuple(self.know_compartments), :]
+        if self.has_teoretic_released:
+            t = tuple([self.configuration_matrix.shape[0] + i for i in self.know_released_compartments])
+            released_target_results = self.last_result.y[t, :]
+        if metric == 'R2':
+            if self.has_teoretic_y:
+                a = np.sum(np.sum(self.w * ((target_results - self.teoretic_y) ** 2), axis=1) / np.sum((self.teoretic_avg - self.teoretic_y) ** 2, axis=1))
+            else:
+                a = 0
+            if self.has_teoretic_released:
+                b = np.sum(np.sum(self.released_w * ((released_target_results - self.teoretic_released) ** 2), axis=1) / np.sum(
+                    (self.released_avg - self.teoretic_released) ** 2, axis=1))
+            else:
+                b = 0
+            return a + b
+        elif metric == 'norm':
+            if self.has_teoretic_y:
+                a = np.linalg.norm(target_results - self.teoretic_y)
+            else:
+                a = 0
+            if self.has_teoretic_released:
+                b = np.linalg.norm(released_target_results - self.teoretic_released)
+            else:
+                b = 0
+            return a + b
+        else:
+            if self.has_teoretic_y:
+                a = np.sum(np.sum(self.w * ((target_results - self.teoretic_y) ** 2), axis=1) / np.sum(
+                    (self.teoretic_avg - self.teoretic_y) ** 2, axis=1))
+            else:
+                a = 0
+            if self.has_teoretic_released:
+                b = np.sum(np.sum(self.released_w * ((released_target_results - self.teoretic_released) ** 2), axis=1) / np.sum(
+                    (self.released_avg - self.teoretic_released) ** 2, axis=1))
+            else:
+                b = 0
+            return a + b
+
+    def optimize(self, method=None, user_method=None, method_is_func=True,
+                 optimization_func_name='__call__', max_step=0.01, **kwargs):
+        x = super().optimize(method, user_method, method_is_func, optimization_func_name, max_step, **kwargs)
+        n = self.configuration_matrix_target_count + self.outputs_target_count + self.volumes_target_count + int(
+            self.need_magic_optimization)
+        if self.released_configuration_matrix_target:
+            self.released_configuration_matrix[self.released_configuration_matrix_target] = x[n:n + self.released_configuration_matrix_target_count]
+        n += self.released_configuration_matrix_target_count
+        if self.released_outputs_target:
+            self.released_outputs[self.released_outputs_target] = x[n: n + self.released_outputs_target_count]
+        n += self.released_outputs_target_count
+        if self.release_parameters_target:
+            self.release_parameters[self.release_parameters_target] = x[n:n + self.release_parameters_target_count]
+        self.released_configuration_matrix_target = None
+        self.released_outputs_target = None
+        return x
+
+    def plot_model(self, compartment_numbers=None, released_compartment_numbers=None,
+                   released_compartment_names=None, compartment_names=None,  left=None, right=None,
+                   y_lims=None, released_y_lims=None, **kwargs):
+        """
+        Функция для построения графиков модели
+
+        Args:
+            compartment_numbers: Камеры, которые нужно отобразить (если не указать, отобразим все)
+            compartment_names: Имена камер
+        """
+        super().plot_model(compartment_numbers=compartment_numbers, compartment_names=compartment_names,
+                           left=left, right=right, y_lims=y_lims, **kwargs)
+
+        if not released_compartment_numbers:
+            released_compartment_numbers = []
+        if not released_compartment_names:
+            released_compartment_names = {}
+        if not released_y_lims:
+            released_y_lims = {}
+        for i in released_compartment_numbers:
+            j = i + self.outputs.size
+            if hasattr(self, "teoretic_x") and hasattr(self, "teoretic_released") and i in self.know_released_compartments:
+                plt.plot(self.teoretic_x, self.teoretic_released[self.know_compartments.index(i)], "*r")
+            plt.plot(self.last_result.t, self.last_result.y[j])
+            plt.title(released_compartment_names.get(i, j))
+            plt.xlim(left=left, right=right)
+            if released_y_lims.get(i):
+                plt.ylim(released_y_lims.get(i))
+            plt.grid()
+            try:
+                plt.show()
+            except AttributeError:
+                plt.savefig(f'{released_compartment_names.get(i, str(j))}.png')
+                plt.cla()
+