我根据这个教程用Python编写了一个3层神经网络,用于玩石头、剪刀、布游戏,样本数据使用-1代表石头,0代表纸,1代表剪刀,以及与教程中类似的数组。我的函数似乎在每次运行时都陷入相对最小值,我正在寻找解决这个问题的办法。程序如下所示。
#math moduleimport numpy as np#sigmoid function converts numbers to percentages(between 0 and 1)def nonlin(x, deriv = False): if (deriv == True): #sigmoid derivative is just return x*(1-x)#output * (output - 1) return 1/(1+np.exp(-x)) #print the sigmoid function#input data: using MOCK RPS DATA, -1:ROCK, 0:PAPER, 1:SCISSORSinput_data = np.array([[1, 1, 1], [0, 0, 0], [-1, -1, -1], [-1, 1, -1]])#also for trainingoutput_data = np.array([[1], [0], [-1], [1]])#random numbers to not get stuck in local minima for fitnessnp.random.seed(1)#create random weights to be trained in loopfirstLayer_weights = 2*np.random.random((3, 4)) - 1 #size of matrixsecondLayer_weights = 2*np.random.random((4, 1)) - 1for value in xrange(60000): # loops through training #pass input through weights to output: three layers layer0 = input_data #layer1 takes dot product of the input and weight matrices, then maps them to sigmoid function layer1 = nonlin(np.dot(layer0, firstLayer_weights)) #layer2 takes dot product of layer1 result and weight matrices, then maps the to sigmoid function layer2 = nonlin(np.dot(layer1, secondLayer_weights)) #check computer predicted result against actual data layer2_error = output_data - layer2 #if value is a factor of 10,000, so six times (out of 60,000), #print how far off the predicted value was from the data if value % 10000 == 0: print "Error:" + str(np.mean(np.abs(layer2_error))) #average error #find out how much to re-adjust weights based on how far off and how confident the estimate layer2_change = layer2_error * nonlin(layer2, deriv = True) #find out how layer1 led to error in layer 2, to attack root of problem layer1_error = layer2_change.dot(secondLayer_weights.T) #^^sends error on layer2 backwards across weights(dividing) to find original error: BACKPROPAGATION #same thing as layer2 change, change based on accuracy and confidence layer1_change = layer1_error * nonlin(layer1, deriv=True) #modify weights based on multiplication of error between two layers secondLayer_weights = secondLayer_weights + layer1.T.dot(layer2_change) firstLayer_weights = firstLayer_weights + layer0.T.dot(layer1_change)
如您所见,这是涉及的数据部分:
input_data = np.array([[1, 1, 1], [0, 0, 0], [-1, -1, -1], [-1, 1, -1]])#also for trainingoutput_data = np.array([[1], [0], [-1], [1]])
权重在这里:
firstLayer_weights = 2*np.random.random((3, 4)) - 1 #size of matrixsecondLayer_weights = 2*np.random.random((4, 1)) - 1
似乎在前一千代之后,权重的调整效率非常低,这让我认为它们已经达到了相对最小值,如下图所示:
有什么快速有效的替代方法可以解决这个问题吗?
回答:
您的网络的一个问题是输出(layer2
元素的值)只能在0到1之间变化,因为您使用了sigmoid非线性。由于您的四个目标值之一是-1,而最接近的可能预测值是0,因此总会至少有25%的误差。以下是一些建议:
-
对输出使用独热编码:即设置三个输出节点——分别对应
石头
、纸
和剪刀
——并训练网络计算这些输出的概率分布(通常使用softmax和交叉熵损失)。 -
将网络的输出层设为线性层(应用权重和偏置,但不使用非线性)。要么添加另一层,要么从当前输出层移除非线性。
您可以尝试的其他方法,但不太可能可靠,因为您实际上是在处理分类数据而不是连续输出:
-
缩放您的数据,使训练数据中的所有输出都在0到1之间。
-
使用产生-1到1之间值的非线性(如tanh)。