diff --git a/深度学习笔记/前馈神经网络/feedforward_neural_network.py b/深度学习笔记/前馈神经网络/feedforward_neural_network.py
index 97a2c71..042c6f0 100755
--- a/深度学习笔记/前馈神经网络/feedforward_neural_network.py
+++ b/深度学习笔记/前馈神经网络/feedforward_neural_network.py
@@ -1,3 +1,4 @@
+import sys
 import numpy as np
 from keras.datasets import imdb
 
@@ -13,16 +14,18 @@ def relu(z):
     '''
     return np.maximum(0, z)
 
-n0, m = X.shape
-n1 = 20
-n2 = 7
-n3 = 5
-n4 = 1
-layers_dims = [n0, n1, n2, n3, n4] # [12288, 20, 7, 5, 1] 
-L = len(layers_dims) - 1  # 4层神经网络，不计输入层
+L = 0
 
 ##### neural network model
-def neural_network(X, Y, learning_rate=0.01, num_iterations=2000, lambd=0):
+def neural_network(X, Y, learning_rate=0.01, num_iterations=8000, lambd=0):
+    n0, m = X.shape
+    n1 = 20
+    n2 = 7
+    n3 = 5
+    n4 = 1
+    layers_dims = [n0, n1, n2, n3, n4] # [12288, 20, 7, 5, 1] 
+    L = len(layers_dims) - 1  # 4层神经网络，不计输入层
+
     m = X.shape[1]
     ### initialize forward propagation
     param_w = [i for i in range(L+1)]
@@ -91,7 +94,13 @@ def neural_network(X, Y, learning_rate=0.01, num_iterations=2000, lambd=0):
         if i % 100 == 0:
             print("cost after iteration {}: {}".format(i, cost))
 
+    parameters = {}
+    parameters["param_w"] = param_w
+    parameters["param_b"] = param_b
+    return parameters
+
 def predict(X_new, parameters, threshold=0.5):
+    L = 4
     param_w = parameters["param_w"]
     param_b = parameters["param_b"]
 
@@ -115,12 +124,22 @@ def vectorize_sequences(sequences, dimension=10000):
         results[i, sequence] = 1.  # 索引results矩阵中的位置，赋值为1，全部都是从第0行0列开始的
     return results
 
-(train_data, train_labels), (test_data, test_labels) = imdb.load_data(num_words=10000)
+(train_data, train_labels), (test_data, test_labels) = imdb.load_data(path="imdb/imdb.npz",num_words=10000)
 
 # Our vectorized training data
-x_train = vectorize_sequences(train_data)
+x_train = vectorize_sequences(train_data)[:1000]
 # Our vectorized test data
-x_test = vectorize_sequences(test_data)
+x_test = vectorize_sequences(test_data)[:5]
+
+y_train = np.asarray(train_labels).astype('float32')[:1000]
+y_test = np.asarray(test_labels).astype('float32')[:5]
+
+parameters = neural_network(x_train.T, y_train)
+y_p = predict(x_test.T, parameters)
+print(y_p)
+print(y_test)
+y_p2 = predict(x_train[:5].T, parameters)
+print(y_p2)
+print(y_train[:5])
 
-y_train = np.asarray(train_labels).astype('float32')
-y_test = np.asarray(test_labels).astype('float32')
\ No newline at end of file
+sys.exit()