[master] v 0.1.1

Author: gcappon

dist/py_replay_bg-0.1.1-py3-none-any.whl

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: py_replay_bg
-Version: 0.1.0
+Version: 0.1.1
 Summary: ReplayBG is a digital twin-based methodology to assess new strategies for type 1 diabetes management.
 Author-email: Giacomo Cappon <cappongiacomo@gmail.com>
 Project-URL: Homepage, https://github.com/gcappon/py_replay_bg
@@ -51,7 +51,7 @@ Requires-Dist: pillow==10.2.0
 Requires-Dist: pkginfo==1.10.0
 Requires-Dist: plotly==5.20.0
 Requires-Dist: pluggy==1.4.0
-Requires-Dist: py-agata==0.0.5
+Requires-Dist: py-agata==0.0.6
 Requires-Dist: Pygments==2.17.2
 Requires-Dist: pyparsing==3.1.2
 Requires-Dist: pyproject_hooks==1.0.0

dist/py_replay_bg-0.1.1.tar.gz

@@ -15,6 +15,7 @@ py_replay_bg/dss/default_dss_handlers.py
 py_replay_bg/environment/__init__.py
 py_replay_bg/example/code/quick-start.py
 py_replay_bg/identification/__init__.py
+py_replay_bg/identification/map.py
 py_replay_bg/identification/mcmc.py
 py_replay_bg/input_validation/__init__.py
 py_replay_bg/model/__init__.py
@@ -25,6 +26,7 @@ py_replay_bg/tests/__init__.py
 py_replay_bg/tests/test_plot_replay_bg_intervals.py
 py_replay_bg/tests/test_replay_bg.py
 py_replay_bg/tests/test_replay_bg_intervals.py
+py_replay_bg/tests/test_replay_bg_intervals_map.py
 py_replay_bg/utils/__init__.py
 py_replay_bg/utils/stats.py
 py_replay_bg/visualizer/__init__.py

py_replay_bg.egg-info/PKG-INFO

@@ -36,7 +36,7 @@ pillow==10.2.0
 pkginfo==1.10.0
 plotly==5.20.0
 pluggy==1.4.0
-py-agata==0.0.5
+py-agata==0.0.6
 Pygments==2.17.2
 pyparsing==3.1.2
 pyproject_hooks==1.0.0

py_replay_bg.egg-info/SOURCES.txt

@@ -123,6 +123,8 @@ def __init__(self, modality, save_name, save_folder, save_suffix='', save_worksp
             os.mkdir(os.path.join(self.replay_bg_path, 'results'))
         if not (os.path.exists(os.path.join(self.replay_bg_path, 'results', 'draws'))):
             os.mkdir(os.path.join(self.replay_bg_path, 'results', 'draws'))
+        if not (os.path.exists(os.path.join(self.replay_bg_path, 'results', 'map'))):
+            os.mkdir(os.path.join(self.replay_bg_path, 'results', 'map'))
         if not (os.path.exists(os.path.join(self.replay_bg_path, 'results', 'logs'))):
             os.mkdir(os.path.join(self.replay_bg_path, 'results', 'logs'))
         if not (os.path.exists(os.path.join(self.replay_bg_path, 'results', 'mcmc_chains'))):

py_replay_bg.egg-info/requires.txt

@@ -0,0 +1,183 @@
+import os
+
+import matplotlib.pyplot as plt
+from matplotlib import pylab
+
+
+import numpy as np
+
+from multiprocessing import Pool
+
+import pickle
+
+import copy
+import pandas as pd
+
+from py_replay_bg.data import ReplayBGData
+
+from datetime import datetime, timedelta
+
+from tqdm import tqdm
+
+from scipy.optimize import minimize
+
+import warnings
+
+# Suppress all RuntimeWarnings
+warnings.filterwarnings("ignore", category=RuntimeWarning)
+
+class MAP:
+    """
+    A class that orchestrates the identification process via MAP.
+    """
+
+    def __init__(self, model,
+                max_iter: int = 100000):
+        """
+        Constructs all the necessary attributes for the MCMC object.
+
+        Parameters
+        ----------
+        model: Model
+            An object that represents the physiological model hyperparameters to be used by ReplayBG.
+        max_iter: int, optional, default : 1000
+            Maximum number of iterations.
+
+        Returns
+        -------
+        None
+
+        Raises
+        ------
+        None
+
+        See Also
+        --------
+        None
+
+        Examples
+        --------
+        None
+        """
+
+        # Physiological model to identify
+        self.model = model
+
+        # Number of unknown parameters to identify
+        self.n_dim = len(self.model.unknown_parameters)
+
+        # Number of times to re-run the procedure
+        self.n_rerun = 64
+
+        # Maximum number of iterations
+        self.max_iter = max_iter
+
+        # Maximum number of function evaluations
+        self.max_fev = 1000000
+
+    def identify(self, data, rbg_data, rbg):
+        """
+        Runs the identification procedure.
+
+        Parameters
+        ----------
+        data: pd.DataFrame
+            Pandas dataframe which contains the data to be used by the tool.
+        rbg_data: ReplayBGData
+            An object containing the data to be used during the identification procedure.
+        rbg: ReplayBG
+            The instance of ReplayBG.
+
+        Returns
+        -------
+        draws: dict
+            A dictionary containing the chain and the samples obtained from the MCMC procedure and the copula sampling, respectively.
+
+        Raises
+        ------
+        None
+
+        See Also
+        --------
+        None
+
+        Examples
+        --------
+        None
+        """
+
+        # Set the initial positions of the walkers.
+        start = self.model.start_guess + self.model.start_guess_sigma * np.random.randn(self.n_rerun, self.n_dim)
+        start[start < 0] = 0
+
+        # Set the pooler
+        pool = None
+        if rbg.environment.parallelize:
+            pool = Pool(processes=rbg.environment.n_processes)
+
+        # Setup the options
+        options = dict()
+        options['maxiter'] = self.max_iter
+        options['maxfev'] = self.max_fev
+        options['disp'] = False
+
+        # Select the function to minimize
+        neg_log_posterior_func = self.model.neg_log_posterior_single_meal if self.model.is_single_meal else self.model.neg_log_posterior_multi_meal
+
+        # Initialize results
+        results = []
+
+        if pool is None:
+
+            if rbg.environment.verbose:
+                iterator = tqdm(range(self.n_rerun))
+                iterator.set_description("Min loss: %f" % np.nan)
+            else:
+                iterator = range(self.n_rerun)
+
+            # Initialize best
+            best = -1
+
+            for r in iterator:
+                result = run_map(start[r], neg_log_posterior_func, rbg_data, options)
+                results.append(result)
+                if best == -1 or result['fun'] < results[best]['fun']:
+                    best = r
+                if rbg.environment.verbose:
+                    iterator.set_description("Min loss %f" % results[best]['fun'])
+
+        else:
+            # Prepare input arguments as tuples for starmap
+            args = [(start[r], neg_log_posterior_func, rbg_data, options) for r in range(self.n_rerun)]
+
+            # Initialize best
+            best = -1
+
+            # Get results (verbosity not allowed for the moment)
+            results = pool.starmap(run_map, args)
+
+            # Get best
+            for r, result in enumerate(results):
+                if best == -1 or result['fun'] < results[best]['fun']:
+                    best = r
+
+        draws = dict()
+        for up in range(len(rbg.model.unknown_parameters)):
+            draws[rbg.model.unknown_parameters[up]] = results[best]['x'][up]
+
+        # Save results
+        identification_results = dict()
+        identification_results['draws'] = draws
+
+        with open(os.path.join(rbg.environment.replay_bg_path, 'results', 'map',
+                               'map_' + rbg.environment.save_name + '.pkl'), 'wb') as file:
+            pickle.dump(identification_results, file)
+
+        return draws
+
+def run_map(start, func, rbg_data, options):
+    result = minimize(func, start, method='Powell', args=(rbg_data,), options=options)
+    ret = dict()
+    ret['fun'] = result.fun
+    ret['x'] = result.x
+    return ret

py_replay_bg/environment/__init__.py

@@ -301,6 +301,19 @@ def validate(self):
                 raise Exception("'previous_data_name' input must be a string.'")
 
 
+class IdentificationMethodValidator:
+    """
+    Class for validating the 'identification_method' input parameter of ReplayBG.
+    """
+
+    def __init__(self, identification_method):
+        self.identification_method = identification_method
+
+    def validate(self):
+        if not (self.identification_method == 'mcmc' or self.identification_method == 'map'):
+            raise Exception("'cgm_model' input must be 'mcmc' or 'map'.")
+
+
 class NStepsValidator:
     """
     Class for validating the 'n_steps' input parameter of ReplayBG.
@@ -595,6 +608,7 @@ def __init__(self, modality, data, bw, u2ss, scenario, save_name, save_suffix,
                  cgm_model,
                  X0,
                  previous_data_name,
+                 identification_method,
                  n_steps, save_chains, analyze_results,
                  CR, CF, GT,
                  meal_generator_handler, meal_generator_handler_params,
@@ -626,6 +640,8 @@ def __init__(self, modality, data, bw, u2ss, scenario, save_name, save_suffix,
         self.X0 = X0
         self.previous_data_name = previous_data_name
 
+        self.identification_method = identification_method
+
         self.n_steps = n_steps
         self.save_chains = save_chains
         self.analyze_results = analyze_results
@@ -709,6 +725,9 @@ def validate(self):
         # Validate the 'previous_data_name' input
         PreviousDataNameValidator(previous_data_name=self.previous_data_name).validate()
 
+        # Validate the 'identification_method' input
+        IdentificationMethodValidator(identification_method=self.identification_method).validate()
+
         # Validate the 'n_steps' input
         NStepsValidator(n_steps=self.n_steps).validate()