Tensorflow.js tf.train.adagrad()函数

Tensorflow.js是谷歌开发的一个开源库，用于在浏览器或节点环境中运行机器学习模型和深度学习神经网络。

tf.train.adagrad()函数用于创建使用自适应梯度算法(adagrad) 的 tf.AdagradOptimizer。

句法：

tf.train.adagrad(learningRate).

参数：

• learningRate：它指定自适应梯度下降算法将使用的学习率。
• initialAccumulatorValue：指定累加器的初始值。它必须是积极的。

返回值：它返回一个 tf.adagradOptimizer。

示例 1：通过学习系数 x、y 来拟合函数f = (x + y)。

Javascript

// importing tensorflow
import tensorflow as tf
  
const xs = tf.tensor1d([0, 1, 2]);
const ys = tf.tensor1d([1.3, 2.5, 3.7]);
  
const x = tf.scalar(Math.random()).variable();
const y = tf.scalar(Math.random()).variable();
  
// Define a function f(x, y) = x + y.
const f = x => x.add(y);
const loss = (pred, label) => 
    pred.sub(label).square().mean();
  
const learningRate = 0.05;
  
// Create adagrad optimizer
const optimizer = 
  tf.train.adagrad(learningRate);
  
// Train the model.
for (let i = 0; i < 5; i++) {
   optimizer.minimize(() => loss(f(xs), ys));
}
  
// Make predictions.
console.log(
`x: ${x.dataSync()}, y: ${y.dataSync()}`);
const preds = f(xs).dataSync();
preds.forEach((pred, i) => {
console.log(`x: ${i}, pred: ${pred}`);
});

输出

x: 0.8561810255050659, y: 0.6922483444213867
x: 0, pred: 0.6922483444213867
x: 1, pred: 1.6922483444213867
x: 2, pred: 2.6922483444213867

示例 2：通过学习系数 a、b、c 来拟合二次函数。

Javascript

// importing tensorflow
import tensorflow as tf
  
const xs = tf.tensor1d([0, 1, 2, 3]);
const ys = tf.tensor1d([1.1, 5.9, 16.8, 33.9]);
  
const a = tf.scalar(Math.random()).variable();
const b = tf.scalar(Math.random()).variable();
const c = tf.scalar(Math.random()).variable();
  
const f = x => a.mul(
  x.square()).add(b.mul(x)).add(c);
const loss = (pred, label) => 
         pred.sub(label).square().mean();
  
const learningRate = 0.01;
const optimizer = 
      tf.train.adagrad(learningRate);
  
// Train the model.
for (let i = 0; i < 10; i++) {
   optimizer.minimize(() => loss(f(xs), ys));
}
  
// Make predictions.
console.log(
`a: ${a.dataSync()}, b: ${b.dataSync()}, c: ${c.dataSync()}`);
const preds = f(xs).dataSync();
preds.forEach((pred, i) => {
   console.log(`x: ${i}, pred: ${pred}`);
});

输出

a: 0.3611285388469696, 
b: 0.6980878114700317, 
c: 0.8787991404533386
x: 0, pred: 0.8787991404533386
x: 1, pred: 1.9380154609680176
x: 2, pred: 3.7194888591766357
x: 3, pred: 6.223219394683838

参考： https://js.tensorflow.org/api/1.0.0/#train.adagrad