Exemplos de make_salad em Python

Exemplo n.º 1

Arquivo: ex_regression.py Projeto: shashankg7/spearmint-salad

'''

from spearmint_salad.high_level import make_salad, format_trace_structure, format_final_risk, get_final_predictions, make_partition
from spearmint_salad.high_level import NormalizeDs

from spearmint_salad import  hp
from sklearn.svm import SVR
hp_space = hp.Obj(SVR)(
    C = hp.Float( 0.01, 1e6, hp.log_scale ),
    gamma = hp.Float( 1e-6, 1e6, hp.log_scale ),
    epsilon = hp.Float(0.01,1, hp.log_scale),
)


from sklearn.datasets.base import load_diabetes, load_boston

dataset_partition = make_partition(load_diabetes, trn_ratio=0.6, val_ratio=0.2)

from spearmint_salad import  metric
metric = metric.SquareDiffLoss()

trace = make_salad( hp_space, metric, dataset_partition, max_iter = 10, salad_size=10)

print format_trace_structure(trace)
print
print format_final_risk(trace)
print

prediction_dict = get_final_predictions(trace)
print 'predictions available for %s.'%(', '.join(prediction_dict.keys()))

Exemplo n.º 2

Arquivo: ex_classification.py Projeto: shashankg7/spearmint-salad

Created on Mar 31, 2014

@author: alex
'''

from spearmint_salad.high_level import make_salad, format_trace_structure, format_final_risk, get_final_predictions, make_partition


from spearmint_salad import  hp
from sklearn.svm import SVC
hp_space = hp.Obj(SVC)(
    C = hp.Float( 0.01, 1000, hp.log_scale ),
    gamma = hp.Float( 1e-7, 1000, hp.log_scale ),
)

from sklearn.datasets.base import load_digits
dataset_partition = make_partition(load_digits, trn_ratio=0.6, val_ratio=0.2)

from spearmint_salad import  metric
metric = metric.ZeroOneLoss()

trace = make_salad( hp_space, metric, dataset_partition, max_iter = 10, mcmc_iters=0)

print format_trace_structure(trace)
print
print format_final_risk(trace)
print

prediction_dict = get_final_predictions(trace)
print 'predictions available for %s.'%(', '.join(prediction_dict.keys()))