纯python代码实现反向传播神经网络

import numpy as np
def sigmod(z):
	h=1./(1+np.exp(-z))
	return h
def de_sigmoid(h):
	return h*(1-h)

# dim in:输入特征的维度
# list_num hidden:每层输出节点的数目# list_actfuns:每层的激活函数
# list de act funs:反向传播时的函数

def bulid_net(dim_in,list_num_hidden,list_act_funs,list_de_act_funs):
	layers=[]
	#逐层的进行网络构建
	for i in range (len(list_num_hidden)):
		layer={}
		#定义每一层的权重if i ==0 :
		layer["w"]=0.2*np.random.randn(dim_in,list_num_hidden[i])-0.1
	else:
		layer["w"]=0.2*np.random.randn(list_num_hidden[i-1],list_num_hidden[i])-0.1
		#定义每一层的偏置
		layer["b"]=0.1*np.ones([1,list_num_hidden[i]])
		layer["act_fun"]=list_act_funs[i]
		layer["de_act_fun"]=list_de_act_funs[i]
		layers.append(layer)
	return layers

返回每一层的输入与最后一层的输出
def fead_forward(datas,layers):
	input_layers=[]
	for i in range(len(layers)):
		layer=layers[i]
		if i ==0:
			inputs=datas
			z=np.dot(inputs,layer["w"])+layer["b"]
			h=layer['act_fun'](z)
			input_layers.append(inputs)
		else:
			inputs=h
			z=np.dot(inputs,layer["w"])+layer["b"]
			h=layer['act_fun'](z)
			input_layers.append(inputs)
	return input_layers,h


进行参数更新更新
def updata_wb(datas,labs,layers, loss_fun,de_loss_fun ,alpha=0.01):
	N,D=np.shape(datas)
	进行前馈操作
	inputs,output=fead_forward(datas,layers)计算loss
	loss=loss_fun(output,labs)
	deltas0=de_loss_fun(output,labs)
	从后向前计算误差deltas=[]
	for i in range(len(layers)):
		index=-i-1
		if i==0:
			delta=deltas0*layers[index]["de_act_fun"](output)
		else:
			h=inputs[index+1]
			delta=np.dot(delta,layers[index+1]["w"].T)*layers[index]["de_act_fun"](h)
		deltas.insert(0,delta)
#利用误差对每一层的权重进行修成for i in range ( len ( layers)) :
#计算dw与db
	for i in range(len(layers)):
		dw=np.dot(inputs[i].T,deltas[i])
		db=np.sum(deltas[i],axis=0,keepdims=True)#梯度下降
		layers[i]["w"]=layers[i]["w"]-alpha*dw
		layers[i]["b"]=layers[i]["b"]-alpha*db
	return layers,loss

def test_accuracy(datas,labs_true,layers) :
	_,output=fead_forward(datas,layers)
	lab_det=np.argmax(output,axis=1)
	labs_true=np.argmax(labs_true,axis=1)
	N_error=np.where(np.abs(labs_true-lab_det)>0)[0].shape[0]
	error_rate=Nerror/np.shape(datas)[0]
	return error rate

def load dataset_iris(file_data,N_train) :
	#数据读取
	datas=np.loadtxt(file_data,dtype=np.float,delimiter=',',usecols=(0,1,2,3))
	labs=np.loadtxt(file_data,dtype=str，delimiter=',',usecols=(4))
	N,D=np.shape(datas)
	N_test =N-N_train
	unqiue_labs =np.unique(labs).tolist()
	dic_str2index={}
	dic_index2str={}
	for i in range(len(unqiue_labs)):
		lab_str=unqiue_labs[i]
		dic_str2index[lab_str]=idic_index2str[i]=lab_str
		labs_onehot=np.zeros([N,len(unqiue_labs)])
	for i in range(N):
		lbs_onehot[i,dic_str2index[labs[i]]]=1
	perm=np.random.permutation(N)
	index_train=perm[:N_train]
	index_test=perm[N_train:]
	data_train=datas[index_train,:]
	lab_train_onehot=labs_onehot[index_train,:]
	data_test=datas[index_test,:]
	lab_test_onehot=labs_onehot[index_test]
	return data_train,lab_train_onehot,data_test,lab_test_onehot,dic_index2str