根据身高推测体重
const $ = require('jquery');
const tf = require('@tensorflow/tfjs');
const tfvis = require('@tensorflow/tfjs-vis');
/* 根据身高推测体重 */
//把数据处理成符合模型要求的格式
function getData() {
//学习数据
const heights = [150, 151, 160, 161, 166, 170, 177, 180, 183, 190, 196];
const weights = [45, 43, 50, 61, 68, 60, 70, 70, 65, 89, 98];
//验证数据
const testh = [183, 160, 151, 166, 183, 177, 170];
const testw = [67, 61, 43, 67, 80, 77, 70];
//归一化数据
const inputs = tf.tensor(heights).sub(150).div(50);
const labels = tf.tensor(weights).sub(40).div(60);
const xs = tf.tensor(testh).sub(150).div(50);
const ys = tf.tensor(testw).sub(40).div(60);
//绘制图表
tfvis.render.scatterplot(
{ name: '身高体重' },
//x轴身高,y轴体重
{ values: heights.map((x, i) => ({ x, y: weights[i] })) },
//设置x轴范围,设置y轴范围
{ xAxisDomain: [140, 200], yAxisDomain: [40, 110] }
);
return { inputs, labels, xs, ys };
}
$(async () => {
const { inputs, labels, xs, ys } = getData();
//设置连续模型
const model = tf.sequential();
//设置全连接层
model.add(tf.layers.dense({
units: 10,
inputShape: [1]
}));
model.add(tf.layers.dense({
units: 1
}));
//设置损失函数,优化函数学习速率为0.1
model.compile({
loss: tf.losses.meanSquaredError,
optimizer: tf.train.adam(0.1)
});
await model.fit(inputs, labels, {
batchSize: 1,
epochs: 20,
//设置验证集
validationData: [xs, ys],
callbacks: tfvis.show.fitCallbacks(
{ name: '训练过程' },
['loss', 'val_loss', 'acc', 'val_acc'],
{ callbacks: ['onEpochEnd'] }
)
});
//对身高180的体重进行推测
let res = model.predict(tf.tensor([180]).sub(150).div(50));
console.log(res.mul(60).add(40).dataSync()[0]);
//保存模型
window.download = async () => {
await model.save('downloads://my-model');
}
});
html部分
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Document</title>
</head>
<body>
<button onclick="download()">保存模型</button>
</body>
<script src="./t1.js"></script>
</html>
运行结果
————————————————
版权声明:本文为CSDN博主「(; ̄ェ ̄)。」的原创文章,遵循CC 4.0 BY-SA版权协议,转载请附上原文出处链接及本声明。
原文链接:https://blog.csdn.net/qq_35496811/article/details/122527040