小编Cha*_*uez的帖子

我试图实现一个前馈神经网络.

这是结构:输入层:8个神经元,隐藏层:8个神经元和输出层:8个神经元.

输入数据是8位的矢量(输入层的每个神经元为1位).神经网络的输出也是8位的向量.所以总共数据集有256个例子.

示例:如果给定x = [0.0,1.0,0.0,0.0,1.0,1.0,0.0,1.0]

输出必须是y = [1.0,0.0,0.0,0.0,0.0,1.0,0.0,1.0]

这是实施:

from keras.models import Sequential
from keras.layers import Dense
import numpy as np
import random
from math import ceil

#Dimension of layers
dim = 8

#Generate dataset
X = []
for i in range(0,2**dim):
    n = [float(x) for x in bin(i)[2:]]
    X.append([0.]*(dim-len(n))+n)
y = X[:]
random.shuffle(y)
X = np.array(X)
y = np.array(y)

# create model
model = Sequential()
model.add(Dense(dim, input_dim=dim, init='normal', activation='sigmoid'))
model.add(Dense(dim, init='normal', activation='sigmoid'))
model.add(Dense(dim, init='normal', activation='sigmoid'))

# Compile model …