写一段python代码：利用keras.models.Sequential()函数，构造一个多层感知器模型。其中，使用Dense()进行全连接层的构造，把MNIST数据集的image 28*28变成784*1 向量；把label从单个数字变成独热向量，使用不同数量的训练数据（5000，10000，15000，…，60000,公差=5000的等差数列，构造输入层784，第一隐层 128,第二隐层 64，输出层10的双隐层感知器，绘制训练集和测试集准确率（纵轴）关于训练数据大小(横轴)的曲线

好的，以下是按照您的要求写的Python代码：

from keras.datasets import mnist
from keras.utils import np_utils
from keras.models import Sequential
from keras.layers import Dense

import numpy as np
import matplotlib.pyplot as plt

# 加载MNIST数据集
(X_train, y_train), (X_test, y_test) = mnist.load_data()

# 将图像数据转换成784*1的向量
X_train = X_train.reshape(X_train.shape[0], 784).astype('float32')
X_test = X_test.reshape(X_test.shape[0], 784).astype('float32')

# 将标签转换成独热向量
y_train = np_utils.to_categorical(y_train)
y_test = np_utils.to_categorical(y_test)

# 构建模型
def build_model():
    model = Sequential()
    model.add(Dense(units=128, input_dim=784, activation='relu'))
    model.add(Dense(units=64, activation='relu'))
    model.add(Dense(units=10, activation='softmax'))
    model.compile(loss='categorical_crossentropy',
                  optimizer='adam', metrics=['accuracy'])
    return model

# 训练模型并绘制准确率曲线
batch_size = 128
epochs = 10

train_sizes = range(5000, 65000, 5000)
train_accs = []
test_accs = []

for size in train_sizes:
    X_train_subset = X_train[:size]
    y_train_subset = y_train[:size]

    model = build_model()
    history = model.fit(X_train_subset, y_train_subset, batch_size=batch_size,
                        epochs=epochs, verbose=0, validation_data=(X_test, y_test))

    train_accs.append(history.history['accuracy'][-1])
    test_accs.append(history.history['val_accuracy'][-1])

plt.plot(train_sizes, train_accs, 'o-', label="Training Accuracy")
plt.plot(train_sizes, test_accs, 'o-', label="Testing Accuracy")
plt.xlabel("Training Set Size")
plt.ylabel("Accuracy")
plt.title("Accuracy vs. Training Set Size")
plt.legend(loc="best")
plt.show()

这段代码会输出一个准确率曲线，其中横轴为训练集大小，纵轴为准确率。您可以根据需要进行修改。