Mercurial > repos > bgruening > create_tool_recommendation_model

comparison optimise_hyperparameters.py @ 2:76251d1ccdcc draft

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
"planemo upload for repository https://github.com/bgruening/galaxytools/tree/recommendation_training/tools/tool_recommendation_model commit 6fa2a0294d615c9f267b766337dca0b2d3637219"
author bgruening
date Fri, 11 Oct 2019 18:24:54 -0400
parents 9bf25dbe00ad
children 5b3c08710e47
comparison
equal deleted inserted replaced
1:12764915e1c5 2:76251d1ccdcc
20 @classmethod 20 @classmethod
21 def __init__(self): 21 def __init__(self):
22 """ Init method. """ 22 """ Init method. """
23 23
24 @classmethod 24 @classmethod
25 def train_model(self, config, reverse_dictionary, train_data, train_labels, test_data, test_labels, class_weights): 25 def train_model(self, config, reverse_dictionary, train_data, train_labels, class_weights):
26 """ 26 """
27 Train a model and report accuracy 27 Train a model and report accuracy
28 """ 28 """
29 l_recurrent_activations = config["activation_recurrent"].split(",") 29 l_recurrent_activations = config["activation_recurrent"].split(",")
30 l_output_activations = config["activation_output"].split(",") 30 l_output_activations = config["activation_output"].split(",")
44 validation_split = float(config["validation_share"]) 44 validation_split = float(config["validation_share"])
45 45
46 # get dimensions 46 # get dimensions
47 dimensions = len(reverse_dictionary) + 1 47 dimensions = len(reverse_dictionary) + 1
48 best_model_params = dict() 48 best_model_params = dict()
49 early_stopping = EarlyStopping(monitor='val_loss', mode='min', min_delta=1e-4, verbose=1, patience=1) 49 early_stopping = EarlyStopping(monitor='val_loss', mode='min', verbose=1, min_delta=1e-4)
50 50
51 # specify the search space for finding the best combination of parameters using Bayesian optimisation 51 # specify the search space for finding the best combination of parameters using Bayesian optimisation
52 params = { 52 params = {
53 "embedding_size": hp.quniform("embedding_size", l_embedding_size[0], l_embedding_size[1], 1), 53 "embedding_size": hp.quniform("embedding_size", l_embedding_size[0], l_embedding_size[1], 1),
54 "units": hp.quniform("units", l_units[0], l_units[1], 1), 54 "units": hp.quniform("units", l_units[0], l_units[1], 1),
80 shuffle="batch", 80 shuffle="batch",
81 verbose=2, 81 verbose=2,
82 validation_split=validation_split, 82 validation_split=validation_split,
83 callbacks=[early_stopping] 83 callbacks=[early_stopping]
84 ) 84 )
85 return {'loss': model_fit.history["val_loss"][-1], 'status': STATUS_OK} 85 return {'loss': model_fit.history["val_loss"][-1], 'status': STATUS_OK, 'model': model}
86 # minimize the objective function using the set of parameters above4 86 # minimize the objective function using the set of parameters above
87 trials = Trials() 87 trials = Trials()
88 learned_params = fmin(create_model, params, trials=trials, algo=tpe.suggest, max_evals=int(config["max_evals"])) 88 learned_params = fmin(create_model, params, trials=trials, algo=tpe.suggest, max_evals=int(config["max_evals"]))
89 print(learned_params) 89 best_model = trials.results[np.argmin([r['loss'] for r in trials.results])]['model']
90
90 # set the best params with respective values 91 # set the best params with respective values
91 for item in learned_params: 92 for item in learned_params:
92 item_val = learned_params[item] 93 item_val = learned_params[item]
93 if item == 'activation_output': 94 if item == 'activation_output':
94 best_model_params[item] = l_output_activations[item_val] 95 best_model_params[item] = l_output_activations[item_val]
95 elif item == 'activation_recurrent': 96 elif item == 'activation_recurrent':
96 best_model_params[item] = l_recurrent_activations[item_val] 97 best_model_params[item] = l_recurrent_activations[item_val]
97 else: 98 else:
98 best_model_params[item] = item_val 99 best_model_params[item] = item_val
99 return best_model_params 100 return best_model_params, best_model