用Tensorflow.js编写数字分类器神经网络

// 导入TensorFlow.js库
import * as tf from '@tensorflow/tfjs';

// 定义神经网络模型
const model = tf.sequential();
model.add(tf.layers.flatten({ inputShape: [28, 28] }));
model.add(tf.layers.dense({ units: 128, activation: 'relu' }));
model.add(tf.layers.dense({ units: 10, activation: 'softmax' }));

// 编译模型
model.compile({ optimizer: 'adam', loss: 'categoricalCrossentropy', metrics: ['accuracy'] });

// 加载训练数据
const trainData = tf.data.csv('train.csv', { columnConfigs: { label: { isLabel: true } } });
const testData = tf.data.csv('test.csv', { columnConfigs: { label: { isLabel: true } } });

// 数据预处理
const preprocessData = (data) => {
  const xs = data.map(({ pixels }) => tf.tensor1d(pixels.split(',').map(Number)));
  const ys = data.map(({ label }) => tf.oneHot(label, 10));
  return { xs, ys };
};

// 训练模型
const trainModel = async () => {
  const { xs, ys } = preprocessData(await trainData.toArray());
  await model.fit(xs, ys, { epochs: 10, batchSize: 32 });
};

// 测试模型
const testModel = async () => {
  const { xs, ys } = preprocessData(await testData.toArray());
  const result = model.evaluate(xs, ys);
  console.log(`Test accuracy: ${result[1].dataSync()[0]}`);
};

// 运行训练和测试
trainModel().then(testModel);