for epoch in range(5):
clf.fit(x_train, y_train, steps=500)
y_pred = clf.predict(x_test)
print('training {}'.format(clf.evaluate(x=x_train, y=y_train)))
print('validation {}'.format(clf.evaluate(x=x_test, y=y_test)))
# print('all {}'.format(clf.evaluate(x=x_all, y=y_all)))
print(metrics.classification_report(y_test, y_pred))
print('confusion matrix\n', metrics.confusion_matrix(y_test, y_pred))
return clf
#clf = TrainLinear(x_train, y_train, x_test, y_test)
clf = TrainLinear(x_all, y_all, x_all, y_all)
clf.get_variable_names()
clf.get_variable_value('linear/_weight/Adagrad')
df_goblin = df[df.type == 'Goblin']
sns.pairplot(df.drop(['id', 'color'], axis=1), hue="type", diag_kind='kde')
sns.pairplot(df_goblin.drop(['id'], axis=1), hue="color", diag_kind='kde')
sns.pairplot(df_goblin.drop(['id'], axis=1), hue="color", diag_kind='hist')
df3 = AddMahalanobis(df)
df3['type'][np.array(id_test)[(y_test == 2)
& (y_pred == 1)]] = 'test_Goblin_pred_Ghoul'
df3['type'][np.array(id_test)[(y_test == 2)
& (y_pred == 0)]] = 'test_Goblin_pred_Ghost'
sns.pairplot(df3.drop(['id'], axis=1), hue="type")
from sklearn import metrics
tf.logging.set_verbosity(tf.logging.ERROR)
