def save_final_model_V3(filename=None, include_position=True):
'''
run_models(produce_final_model=True) is what saves the model
'''
test = False
assert filename is not None, "need to provide filename to save final model"
if include_position:
learn_options = {
"V": 3,
'train_genes': load_data.get_V3_genes(),
'test_genes': load_data.get_V3_genes(),
"testing_non_binary_target_name": 'ranks',
'include_pi_nuc_feat': True,
"gc_features": True,
"nuc_features": True,
"include_gene_position": True,
"include_NGGX_interaction": True,
"include_Tm": True,
"include_strand": False,
"include_gene_feature": False,
"include_gene_guide_feature": 0,
"extra pairs": False,
"weighted": None,
"training_metric": 'spearmanr',
"NDGC_k": 10,
"cv": "gene",
"include_gene_effect": False,
"include_drug": False,
"include_sgRNAscore": False,
'adaboost_loss':
'ls', # main "ls", alternatives: "lad", "huber", "quantile", see scikit docs for details
'adaboost_alpha':
0.5, # this parameter is only used by the huber and quantile loss functions.
'normalize_features': False,
}
else:
learn_options = {
"V": 3,
'train_genes': load_data.get_V3_genes(),
'test_genes': load_data.get_V3_genes(),
"testing_non_binary_target_name": 'ranks',
'include_pi_nuc_feat': True,
"gc_features": True,
"nuc_features": True,
"include_gene_position": False,
"include_NGGX_interaction": True,
"include_Tm": True,
"include_strand": False,
"include_gene_feature": False,
"include_gene_guide_feature": 0,
"extra pairs": False,
"weighted": None,
"training_metric": 'spearmanr',
"NDGC_k": 10,
"cv": "gene",
"include_gene_effect": False,
"include_drug": False,
"include_sgRNAscore": False,
'adaboost_loss':
'ls', # main "ls", alternatives: "lad", "huber", "quantile", see scikit docs for details
'adaboost_alpha':
0.5, # this parameter is only used by the huber and quantile loss functions.
'normalize_features': False,
}
learn_options_set = {'final': learn_options}
results, all_learn_options = run_models(
["AdaBoost"],
orders=[2],
adaboost_learning_rates=[0.1],
adaboost_max_depths=[3],
adaboost_num_estimators=[100],
adaboost_CV=False,
learn_options_set=learn_options_set,
test=test,
CV=False)
model = results.values()[0][3][0]
with open(filename, 'wb') as f:
pickle.dump((model, learn_options), f, -1)
return model
def save_final_model_V3(filename=None, include_position=True):
'''
run_models(produce_final_model=True) is what saves the model
'''
test = False
assert filename is not None, "need to provide filename to save final model"
if include_position:
learn_options = {"V": 3,
'train_genes': load_data.get_V3_genes(), 'test_genes': load_data.get_V3_genes(),
"testing_non_binary_target_name": 'ranks',
'include_pi_nuc_feat': True,
"gc_features": True,
"nuc_features": True,
"include_gene_position": True,
"include_NGGX_interaction": True,
"include_Tm": True,
"include_strand": False,
"include_gene_feature": False,
"include_gene_guide_feature": 0,
"extra pairs": False,
"weighted": None,
"training_metric": 'spearmanr',
"NDGC_k": 10,
"cv": "gene",
"include_gene_effect": False,
"include_drug": False,
"include_sgRNAscore": False,
'adaboost_loss' : 'ls', # main "ls", alternatives: "lad", "huber", "quantile", see scikit docs for details
'adaboost_alpha': 0.5, # this parameter is only used by the huber and quantile loss functions.
'normalize_features': False,
}
else:
learn_options = {"V": 3,
'train_genes': load_data.get_V3_genes(), 'test_genes': load_data.get_V3_genes(),
"testing_non_binary_target_name": 'ranks',
'include_pi_nuc_feat': True,
"gc_features": True,
"nuc_features": True,
"include_gene_position": False,
"include_NGGX_interaction": True,
"include_Tm": True,
"include_strand": False,
"include_gene_feature": False,
"include_gene_guide_feature": 0,
"extra pairs": False,
"weighted": None,
"training_metric": 'spearmanr',
"NDGC_k": 10,
"cv": "gene",
"include_gene_effect": False,
"include_drug": False,
"include_sgRNAscore": False,
'adaboost_loss' : 'ls', # main "ls", alternatives: "lad", "huber", "quantile", see scikit docs for details
'adaboost_alpha': 0.5, # this parameter is only used by the huber and quantile loss functions.
'normalize_features': False,
}
learn_options_set = {'final': learn_options}
results, all_learn_options = run_models(["AdaBoost"], orders=[2], adaboost_learning_rates=[0.1],
adaboost_max_depths=[3], adaboost_num_estimators=[100],
adaboost_CV=False, learn_options_set=learn_options_set,
test=test, CV=False)
model = results.values()[0][3][0]
with open(filename, 'wb') as f:
pickle.dump((model, learn_options), f, -1)
return model
orders=[1],
adaboost_learning_rates=[0.1],
adaboost_max_depths=[3],
adaboost_num_estimators=[100],
adaboost_CV=False,
learn_options_set=learn_options_set,
test=test,
produce_final_model=True)
#results, all_learn_options = run_models([ "AdaBoost"], orders=[2], adaboost_learning_rates=[0.1], adaboost_max_depths=[3], adaboost_num_estimators=[100], adaboost_CV=False, learn_options_set=learn_options_set, test=test)
# all_metrics, gene_names = util.get_all_metrics(results, test_metrics=['AUC', 'spearmanr'], learn_options_set=learn_options_set)
# util.plot_all_metrics(all_metrics, gene_names, all_learn_options, save=True)
# plt.close('all')
else:
learn_options = {
"V": 3,
"train_genes": load_data.get_V3_genes(),
"test_genes": load_data.get_V3_genes(),
"testing_non_binary_target_name": 'ranks',
'include_pi_nuc_feat': True,
"gc_features": True,
"nuc_features": True,
"include_gene_position": True,
"include_NGGX_interaction": True,
"include_Tm": True,
"include_strand": False,
"include_gene_feature": False,
"include_gene_guide_feature": 0,
"extra pairs": False,
"weighted": None,
"training_metric": 'spearmanr',
"NDGC_k": 10,
thiskey = 'try_models'
learn_options_set = {thiskey: learn_options_2}
print "working on %s" % thiskey
#learn_options_set = {'drug_gene': learn_options, 'drug': learn_options_2}
#results, all_learn_options = run_models(['linreg', 'L1', 'L2'], orders=[1], test=True, target_name='score')
# all_learn_options is similar to (and contains all of_ learn_options_set, but has more
# entries populated, by the model specification, etc.
results, all_learn_options = run_models(["AdaBoost"], orders=[1], adaboost_learning_rates=[0.1], adaboost_max_depths=[3], adaboost_num_estimators=[100], adaboost_CV=False, learn_options_set=learn_options_set, test=test, produce_final_model=True)
#results, all_learn_options = run_models([ "AdaBoost"], orders=[2], adaboost_learning_rates=[0.1], adaboost_max_depths=[3], adaboost_num_estimators=[100], adaboost_CV=False, learn_options_set=learn_options_set, test=test)
# all_metrics, gene_names = util.get_all_metrics(results, test_metrics=['AUC', 'spearmanr'], learn_options_set=learn_options_set)
# util.plot_all_metrics(all_metrics, gene_names, all_learn_options, save=True)
# plt.close('all')
else :
learn_options = {"V": 3,
"train_genes": load_data.get_V3_genes(),
"test_genes": load_data.get_V3_genes(),
"testing_non_binary_target_name": 'ranks',
'include_pi_nuc_feat': True,
"gc_features": True,
"nuc_features": True,
"include_gene_position": True,
"include_NGGX_interaction": True,
"include_Tm": True,
"include_strand": False,
"include_gene_feature": False,
"include_gene_guide_feature": 0,
"extra pairs": False,
"weighted": None,
"training_metric": 'spearmanr',
"NDGC_k": 10,
