grid_search.py

import sklearn.model_selection as model_selection

import _config_grid_search

import numpy as np
import search

import database

class GridSearch(search.Search):
    def __init__(self, estimator, chain_names, hyperparameter_space,
                 dataset_name, verbose=0):

        super(GridSearch, self).__init__()
        
        self.cv = _config_grid_search.cv_folds
        self.logarithmic_scale = _config_grid_search.logarithmic_scale
        self.chain_names = chain_names
                
        self.gs = model_selection.GridSearchCV(                 
                 estimator=estimator,
                 param_grid = hyperparameter_space,
                 error_score=np.NaN,                 
                 cv=self.cv,                                  
                 verbose=verbose,
                 refit=False,
                 n_jobs=1)      
        
        self.estimator = estimator
        self.dataset_name = dataset_name   
           

    def get_name(self):
        if self.logarithmic_scale == True:
            return "Grid Search (scikit-learn) log"      
        else:
            return "Grid Search (scikit-learn)"
                  
    def fit(self, X, y):
               
        self.gs.fit(X, y)
        
        # write results to the file
        f, file_name = self.results_file_open("w")                        
        
        self.write_header(f)
        if self.logarithmic_scale == True:
            f.write("# logarithmic_scale")
                                                                                                                          
        f.write("# Grid scores on development set:\n\n")

        means = self.gs.cv_results_['mean_test_score']
        stds = self.gs.cv_results_['std_test_score']
        results = []
        for mean, std, params in zip(means, stds, self.gs.cv_results_['params']):
            f.write("%0.3f (+/-%0.03f) for %r"
                  % (mean, std * 2, params))
            f.write("\n")

            # write results to database
            results.append((self.experiment_id, self.get_name(), str(self.chain_names[:-1]), str(self.chain_names[-1]), str(params), self.cv, self.dataset_name, float(mean), float(std), 0))                                          

        f.write("\n\n# Grid Search finished.")
        f.close()
                                                    
        try:
            database.insert_results(results)
        except Exception as e:
            # print(results)
            print(e)                            

        print("Grid Search finished.")