网站开发众包平台,ps软件下载2022,东莞网站se,个体工商户如何注销利用4层神经网络实现小猫的分类#xff0c;小猫训练样本是#xff08;209#xff0c;64*64*312288#xff09;,故输入节点是12288个#xff0c;隐藏层节点依次为20#xff0c;7#xff0c;5#xff0c;输出层为1。 首先看文件路径#xff0c;dnn_utils_v2.py代码是激活…利用4层神经网络实现小猫的分类小猫训练样本是20964*64*312288,故输入节点是12288个隐藏层节点依次为2075输出层为1。 首先看文件路径dnn_utils_v2.py代码是激活函数和激活函数导数 载入数据集 打印预测错误照片的代码 import numpy as np
import matplotlib.pyplot as plt
import h5pydef sigmoid(Z):Implements the sigmoid activation in numpyArguments:Z -- numpy array of any shapeReturns:A -- output of sigmoid(z), same shape as Zcache -- returns Z as well, useful during backpropagationA 1/(1np.exp(-Z))cache Zreturn A, cachedef relu(Z):Implement the RELU function.Arguments:Z -- Output of the linear layer, of any shapeReturns:A -- Post-activation parameter, of the same shape as Zcache -- a python dictionary containing A ; stored for computing the backward pass efficientlyA np.maximum(0,Z)assert(A.shape Z.shape)cache Z return A, cachedef relu_backward(dA, cache):Implement the backward propagation for a single RELU unit.Arguments:dA -- post-activation gradient, of any shapecache -- Z where we store for computing backward propagation efficientlyReturns:dZ -- Gradient of the cost with respect to ZZ cachedZ np.array(dA, copyTrue) # just converting dz to a correct object.# When z 0, you should set dz to 0 as well. dZ[Z 0] 0assert (dZ.shape Z.shape)return dZdef sigmoid_backward(dA, cache):Implement the backward propagation for a single SIGMOID unit.Arguments:dA -- post-activation gradient, of any shapecache -- Z where we store for computing backward propagation efficientlyReturns:dZ -- Gradient of the cost with respect to ZZ caches 1/(1np.exp(-Z))dZ dA * s * (1-s)assert (dZ.shape Z.shape)return dZ
载入数据集def load_data():train_dataset h5py.File(datasets/train_catvnoncat.h5, r)train_set_x_orig np.array(train_dataset[train_set_x][:]) # your train set featurestrain_set_y_orig np.array(train_dataset[train_set_y][:]) # your train set labelstest_dataset h5py.File(datasets/test_catvnoncat.h5, r)test_set_x_orig np.array(test_dataset[test_set_x][:]) # your test set featurestest_set_y_orig np.array(test_dataset[test_set_y][:]) # your test set labelsclasses np.array(test_dataset[list_classes][:]) # the list of classestrain_set_y_orig train_set_y_orig.reshape((1, train_set_y_orig.shape[0]))test_set_y_orig test_set_y_orig.reshape((1, test_set_y_orig.shape[0]))return train_set_x_orig, train_set_y_orig, test_set_x_orig, test_set_y_orig, classesdef print_mislabeled_images(classes, X, y, p):Plots images where predictions and truth were different.X -- datasety -- true labelsp -- predictionsa p ymislabeled_indices np.asarray(np.where(a 1))plt.rcParams[figure.figsize] (40.0, 40.0) # set default size of plotsnum_images len(mislabeled_indices[0])for i in range(num_images):index mislabeled_indices[1][i]plt.subplot(2, num_images, i 1)plt.imshow(X[:, index].reshape(64, 64, 3), interpolationnearest)plt.axis(off)plt.title(Prediction: classes[int(p[0, index])].decode(utf-8) \n Class: classes[y[0, index]].decode(utf-8))
StartDeepNeural.py是整个模型前向传播和后向传播的代码 import numpy as np
import dnn_utils_v2
def initialize_parameters(n_x,n_h,n_y):W1 np.random.randn(n_h,n_x)*0.01b1 np.zeros((n_h,1))W2 np.random.randn(n_y, n_h) * 0.01b2 np.zeros((n_y,1))assert (W1.shape(n_h,n_x))assert (b1.shape (n_h, 1))assert (W2.shape (n_y, n_h))assert (b2.shape (n_y, 1))parameters{W1: W1,b1: b1,W2: W2,b2: b2}return parameters多层神经网络的参数初始化注意权重的初始化 为防止梯度爆炸和梯度消失def initialize_parameters_deep(layer_dims):Llen(layer_dims)parameters{}for i in range(1,L):parameters[Wstr(i)] np.random.randn(layer_dims[i],layer_dims[i-1])*np.sqrt(2.0/layer_dims[i-1])parameters[b str(i)] np.zeros((layer_dims[i], 1))assert (parameters[Wstr(i)].shape(layer_dims[i],layer_dims[i-1]))assert (parameters[b str(i)].shape (layer_dims[i], 1))return parameters前向传播过程中某一层未加激活函数的操作def linear_forward(A,W,b):Z np.dot(W,A)bassert(Z.shape(W.shape[0],A.shape[1]))cache(A,W,b)return Z,cache前向传播过程中 某一层通过激活函数来采取相应的操作def linear_activation_forward(A_prev,W,b,activation):if activationsigmoid:Z, linear_cachelinear_forward(A_prev,W,b)A, activation_cache dnn_utils_v2.sigmoid(Z)elif activationrelu:Z, linear_cache linear_forward(A_prev, W, b)A, activation_cache dnn_utils_v2.relu(Z)assert(A.shape(W.shape[0],A_prev.shape[1]))cache(linear_cache,activation_cache)###save ( (A W b ),Z) tuplereturn A,cache整个模型的前向传播过程def L_model_forward(X,parameters):Llen(parameters)//2AXcaches[]for i in range(1,L):A_prevAA, cachelinear_activation_forward(A_prev,parameters[Wstr(i)],parameters[bstr(i)],activationrelu)caches.append(cache)AL, cache linear_activation_forward(A, parameters[W str(L)], parameters[b str(L)],activationsigmoid)caches.append(cache)assert(AL.shape(parameters[W str(L)].shape[0],X.shape[1]))return AL,caches计算损失值def compute_cost(AL,Y):m Y.shape[1]cost (1. / m) * (-np.dot(Y, np.log(AL).T) - np.dot(1 - Y, np.log(1 - AL).T))#cost -1 / m * (np.dot(Y, np.log(AL).T) np.dot((1 - Y), np.log(1 - AL).T))cost np.squeeze(cost)assert (cost.shape())return cost后向传播过程中某一层未加激活函数的操作def linear_backward(dZ,cache):A_prev,W,bcachemA_prev.shape[1]dW1/m*np.dot(dZ,A_prev.T)db1/m*np.sum(dZ,axis1,keepdimsTrue)dA_prevnp.dot(W.T,dZ)assert (dW.shape W.shape)assert (db.shape b.shape)assert (dA_prev.shape A_prev.shape)return dA_prev,dW,db后向传播过程中 某一层通过激活函数来采取相应的操作def linear_activation_backward(dA,cache,activation):linear_cache, activation_cachecacheif activationrelu:dZdnn_utils_v2.relu_backward(dA, activation_cache)dA_prev, dW, dblinear_backward(dZ, linear_cache)elif activationsigmoid:dZdnn_utils_v2.sigmoid_backward(dA, activation_cache)dA_prev, dW, db linear_backward(dZ, linear_cache)return dA_prev, dW, db整个模型的后向传播过程def L_model_backward(AL,Y,caches):grads{}Llen(caches) ###[ [( (X W1 b1 ),Z1)],[( (A1 W2 b2 ),Z2) ] [( (A2 W3 b3 ),Z3) ]]mAL.shape[1]YY.reshape(AL.shape)dAL-np.divide(Y,AL)np.divide((1-Y),(1-AL))current_cachecaches[L-1]grads[dAstr(L)],grads[dWstr(L)],grads[dbstr(L)]linear_activation_backward(dAL, current_cache, activationsigmoid)for i in reversed(range(L-1)):current_cache caches[i]grads[dA str(i1)], grads[dW str(i1)], grads[db str(i1)]linear_activation_backward(grads[dAstr(i2)], current_cache, activationrelu)return grads更新参数def update_parameters(parameters,grads,learning_rate):Llen(parameters)//2for i in range(L):parameters[Wstr(i1)]parameters[Wstr(i1)]-learning_rate*grads[dW str(i1)]parameters[b str(i 1)] parameters[b str(i 1)] - learning_rate * grads[db str(i 1)]return parameters DeepNeuralCat_noCat.py就是训练的代码首先看数据集 import numpy as np
import dnn_utils_v2
import matplotlib.pyplot as plt
import StartDeepNeural
train_x_orig, train_y_orig, test_x_orig, test_y_orig,classadnn_utils_v2.load_data()
print(train{}.format(train_x_orig.shape))
print(train_y_orig.shape)
print(train_y_orig[:,:10])
print(test_x_orig.shape)
print(test_y_orig.shape)
print(classa)
plt.imshow(train_x_orig[0])
plt.show() 打印结果 可知训练样本为209个维度为64643测试样本为50个维度为646430代表不是猫 拉成二维向量 对于训练样本20964*64*3,测试样本(50,64*64*3) import numpy as np
import dnn_utils_v2
import matplotlib.pyplot as plt
import StartDeepNeural
train_x_orig, train_y_orig, test_x_orig, test_y_orig,classadnn_utils_v2.load_data()
# print(train{}.format(train_x_orig.shape))
# print(train_y_orig.shape)
# print(train_y_orig[:,:10])
# print(test_x_orig.shape)
# print(test_y_orig.shape)
# print(classa)
# plt.imshow(train_x_orig[0])
# plt.show()
##train
train_x_flatten train_x_orig.reshape(train_x_orig.shape[0],train_x_orig.shape[1] *train_x_orig.shape[2] * 3).T
train_x train_x_flatten / 255.
#print(train_x.shape)
##test
test_x_flatten test_x_orig.reshape(test_x_orig.shape[0],test_x_orig.shape[1] *test_x_orig.shape[2] * 3).T
test_x test_x_flatten / 255.
#print(test_x.shape)开始训练下面代码也写了两层神经网络只是没有用到而已。 import numpy as np
import dnn_utils_v2
import matplotlib.pyplot as plt
import StartDeepNeural
train_x_orig, train_y_orig, test_x_orig, test_y_orig,classadnn_utils_v2.load_data()
# print(train{}.format(train_x_orig.shape))
# print(train_y_orig.shape)
# print(train_y_orig[:,:10])
# print(test_x_orig.shape)
# print(test_y_orig.shape)
# print(classa)
# plt.imshow(train_x_orig[0])
# plt.show()
##train
train_x_flatten train_x_orig.reshape(train_x_orig.shape[0],train_x_orig.shape[1] *train_x_orig.shape[2] * 3).T
train_x train_x_flatten / 255.
#print(train_x.shape)
##test
test_x_flatten test_x_orig.reshape(test_x_orig.shape[0],test_x_orig.shape[1] *test_x_orig.shape[2] * 3).T
test_x test_x_flatten / 255.
#print(test_x.shape)def two_layer_model(X,Y,num_iterations,learning_rate):n_x 12288n_h 7n_y 1layer_dims (n_x, n_h, n_y)#mX.shape[1]#(n_x, n_h, n_y)layer_dimsparametersStartDeepNeural.initialize_parameters_deep(layer_dims)# W1 parameters[W1]# W2 parameters[W2]# b1 parameters[b1]# b2 parameters[b2]costs[]#return W1,W2,b1,b2for i in range(0,num_iterations):# A1, cache1StartDeepNeural.linear_activation_forward(X,W1,b1,activationrelu)# A2, cache2 StartDeepNeural.linear_activation_forward(A1, W2, b2, activationsigmoid)AL, cachesStartDeepNeural.L_model_forward(X, parameters) ##[ [( (X W1 b1 ),Z1)],[( (A1 W2 b2 ),Z2) ]]costStartDeepNeural.compute_cost(AL, Y)gradsStartDeepNeural.L_model_backward(AL, Y, caches)parametersStartDeepNeural.update_paremeters(parameters, grads, learning_rate)if i %1000:print(iterations{}:cost {}.format(i,cost))costs.append(cost)return costs,parameters
def predict(X,Y,parameters):AL, caches StartDeepNeural.L_model_forward(X, parameters)##AL.shape(1,m)mX.shape[1]pnp.zeros((1,m))for i in range(AL.shape[1]):if AL[0][i]0.5:p[0][i]1else:p[0][i] 0result np.squeeze(np.dot(p, Y.T) np.dot(1 - p, 1 - Y.T))accuracyresult/mreturn accuracy,p
def two_layer_model_test():costs, parameters two_layer_model(train_x, train_y_orig,num_iterations 3000, learning_rate 0.0075)# print(parameters)accuracy ,p_train predict(train_x, train_y_orig, parameters)print(train accuracy is {}.format(accuracy))accuracy ,p_test predict(test_x, test_y_orig, parameters)print(test accuracy is {}.format(accuracy))plt.plot(costs)plt.xlabel(iterations)plt.ylabel(costs)plt.title(learning rate is 0.0075)plt.show()def L_layer_model(X, Y, layer_dims,learning_rate0.0075,num_iterations2000):parameters StartDeepNeural.initialize_parameters_deep(layer_dims)costs []# return W1,W2,b1,b2for i in range(0, num_iterations):# A1, cache1StartDeepNeural.linear_activation_forward(X,W1,b1,activationrelu)# A2, cache2 StartDeepNeural.linear_activation_forward(A1, W2, b2, activationsigmoid)AL, caches StartDeepNeural.L_model_forward(X, parameters) ##[ [( (X W1 b1 ),Z1)],[( (A1 W2 b2 ),Z2) ]]cost StartDeepNeural.compute_cost(AL, Y)grads StartDeepNeural.L_model_backward(AL, Y, caches)parameters StartDeepNeural.update_parameters(parameters, grads, learning_rate)if i % 100 0:#print(grads)print(iterations{}:cost {}.format(i, cost))costs.append(cost)return costs, parameters
def print_mislabel_images(classes,test_x,test_y_orig,p_test):atest_y_origp_testmislable_indexnp.asarray(np.where(a1))#[[1,0]] [[1,1]]np.where 返回array([0]),array([1])##np.asarray 返回 array([[0],[1]])plt.figure(figsize(40, 40)) # set default size of plots 画布大小 4000×4000num_imageslen(mislable_index[0])for i in range(num_images):indexmislable_index[1][i]plt.subplot(2,num_images,i1)plt.imshow(test_x[:,index].reshape(64,64,3))#plt.axis(off)plt.title(prediction classes[int(p_test[0][index])].decode(utf-8) real classes[int(test_y_orig[0][index])].decode(utf-8))plt.savefig(1.jpg)
def L_layer_model_test():layers_dims [12288, 20, 7, 5, 1]costs,parameters L_layer_model(train_x, train_y_orig, layers_dims)accuracy , p_train predict(train_x, train_y_orig, parameters)print(train accuracy is {}.format(accuracy))accuracy , p_test predict(test_x, test_y_orig, parameters)print(test accuracy is {}.format(accuracy))print_mislabel_images(classa,test_x,test_y_orig,p_test)# plt.plot(costs)# plt.xlabel(iterations)# plt.ylabel(costs)# plt.title(learning rate is 0.0075)plt.show()
if __name____main__:L_layer_model_test() 打印结果
