如何仅预测标签而不是概率答案

【问题标题】：How to predict label only instead of probability如何仅预测标签而不是概率
【发布时间】：2018-07-20 07:15:32
【问题描述】：

当我预测输出时，我得到的是每个标签的概率（示例输出 = [0.002345, 0.997655]）而不是正确的标签。有没有办法只获得标签而不是他们的概率？

import tflearn    
from tflearn.layers.core import input_data, dropout, fully_connected
from tflearn.layers.estimator import regression
from tflearn.data_utils import load_csv
import numpy as np

data, labels = load_csv("C:/Users/Animkush/Desktop/train.csv", 
categorical_labels=True, n_classes=2, target_column=30)


def preprocess(data, col):
    for id in sorted(col, reverse=True):
    [r.pop(id) for r in data]

    return np.array(data, dtype=np.float32)

ignore = [0]

data = preprocess(data, ignore)


network = input_data(shape=[None, 28], name="input")

network = tflearn.fully_connected(network, 32, activation="relu")
network = tflearn.fully_connected(network, 64, activation="relu")
network = tflearn.fully_connected(network, 128, activation="relu")
network = tflearn.fully_connected(network, 64, activation="relu")
network = tflearn.fully_connected(network, 32, activation="relu")

network = dropout(network, 0.8)

network = fully_connected(network, 2, activation='softmax')
network = regression(network, optimizer='adam', learning_rate=0.001, loss='categorical_crossentropy', name='targets')

model = tflearn.DNN(network)

model.fit(data, labels, n_epoch=100, batch_size=500, show_metric=True)

p = np.array([[-3.043540624,-3.157307121,1.08846278,2.288643618,1.35980513,-1.064822523,0.325574266,-0.067793653,-0.270952836,-0.838586565,-0.414575448,-0.50314086,0.676501545,-1.692028933,2.000634839,0.666779696,0.599717414,1.725321007,0.28334483,2.102338793,0.661695925,0.435477209,1.375965743,-0.293803153,0.279798032,-0.145361715,-0.252773123,0.035764225]])

print(model.predict(p))

【问题讨论】：

我不确定我是否遵循，但有一个层将机器的结果映射到对人类更有意义的东西可能是有意义的......

标签： python machine-learning neural-network tflearn

【解决方案1】：

Numpy 方法argmax 经常用于此目的：

import numpy as np

p = np.array([0.002345, 0.997655]) # array of probabilities, as per your example
y = np.argmax(p) # hard class
y
# 1

# another example:
p = np.array([0.78, 0.22])
y = np.argmax(p)
y
# 0

【讨论】：