def __init__(self):
# config
self.cfg = config.Configuration()
# ROC params
self.__truePos = 0
self.__falsePos = 0
self.__trueNeg = 0
self.__falseNeg = 0
self.__rec = [] # recall
self.__far = [] # false alarm rate
def __init__(self):
cfg = config.Configuration()
self.mlp = MLP(cfg.mlp_n_in, cfg.mlp_n_out)
self.n_epochs = cfg.mlp_n_epochs
self.patience = cfg.mlp_patience
self.learning_rate = cfg.mlp_learning_rate
self.batch_size = cfg.mlp_batch_size
self.l2_reg = cfg.mlp_l2_reg
self.improvement_threshold = cfg.mlp_improvement_threshold
self.patience_increase = cfg.mlp_patience_increase
# load data
dataset = 'mnist.pkl.gz'
datasets = load_data(dataset)
self.train_set_x, self.train_set_y = datasets[0]
self.valid_set_x, self.valid_set_y = datasets[1]
self.test_set_x, self.test_set_y = datasets[2]
def __init__(self, n_in, n_out):
self.cfg = config.Configuration()
self.n_class = n_out
self.W = np.zeros((n_in, n_out), dtype=np.float)
self.b = np.zeros((n_out, ), dtype=np.float)
self.learning_rate = self.cfg.lr_learning_rate
self.n_epochs = self.cfg.lr_n_epoch
self.batch_size = self.cfg.lr_batch_size
# compute p_y given x
self.p_y_given_x = None
self.exp_x_multiply_w_plus_b = None
# gradient_w_b
self.delta_W = 0
self.delta_b = 0
# train
self.patience = self.cfg.lr_patience
self.patience_increase = self.cfg.lr_patience_increase
self.improvement_threshold = self.cfg.lr_improvement_threshold
self.lamda = self.cfg.lr_weight_decay_lamda
# data set
self.train_set_x = None
self.train_set_y = None
self.valid_set_x = None
self.valid_set_y = None
self.test_set_x = None
self.test_set_y = None
# params
self.is_weight_decay = self.cfg.lr_weight_decay
self.is_line_search = self.cfg.lr_line_search
self.is_momentum = self.cfg.is_momentum
self.momentum_vW = 0
self.momentum_vb = 0
# line search test
self.num_test_epoch = 200
self.test_result = np.zeros((self.num_test_epoch + 1, 4),
dtype=np.float32)
plt.legend(loc='upper right')
plt.show()
def printEval(self):
print("Threshold = {:.2f}".format(self.__classifier.getThreshold()))
print("True Positive = {:}".format(self.__truePos))
print("False Negtive = {:}".format(self.__falseNeg))
print("True Negtive = {:}".format(self.__trueNeg))
print("False Positive = {:}".format(self.__falsePos))
self.__calculate()
print("Accuracy = {:.2f}%".format(self.accuracy))
print("Precision = {:.2f}%".format(self.precision))
print("Recall = {:.2f}%".format(self.recall))
print("False alarm rate = {:.2f}%".format(self.falseAlarmRate))
print()
if __name__ == '__main__':
cfg = config.Configuration()
if cfg.task == 'train':
########## train the classifier ############
###### neg as center
classifier = Classifier(cfg)
########## testROC ############################
myTest = TestROC(cfg)
myTest.setClassifier(classifier)
myTest.compute()
myTest.draw()