Tensorflow.js tf.train.adadelta()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。
我们用来创建使用 adadelta 算法的 tf.AdadeltaOptimizer 的tf.train.adadelta()函数。 adadelta算法是梯度体面优化算法的扩展。它用于优化神经网络。
句法:
tf.train.adadelta(learningRate)
参数:
- learningRate:它指定adadelta梯度下降算法将使用的学习率。
- rho:它指定每次更新的学习率衰减。
- epsilon:它指定一个常数 epsilon,用于改善 grad update 的条件。可选的
返回值:它返回一个 tf.adadeltaOptimizer
示例 1:通过学习系数 a 和 b,使用 adadelta 优化器拟合函数f=(a*x+y)。
Javascript
// Importing tensorflow
import * as tf from "@tensorflow/tfjs"
const xs = tf.tensor1d([0, 1, 2, 3]);
const ys = tf.tensor1d([1.1, 5.9, 16.8, 33.9]);
// Choosing variable coefficients
const a = tf.scalar(Math.random()).variable();
const b = tf.scalar(Math.random()).variable();
// Defining function f = (a*x + b)
const f = x => a.mul(x).add(b);
const loss = (pred, label) => pred.sub(label).square().mean();
const learningRate = 0.01;
// Creating optimizer
const optimizer = tf.train.adadelta(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()}}`);
const preds = f(xs).dataSync();
preds.forEach((pred, i) => {
console.log(`x: ${i}, pred: ${pred}`);
});
Javascript
// Importing tensorflow
import * as tf from "@tensorflow/tfjs"
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();
// Setting configurations for our optimizer
const learningRate = 0.01;
const rho = 0.2;
const epsilon = 0.5;
// Creating the optimizer
const optimizer = tf.train.adadelta(learningRate, rho, epsilon);
// 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: 5.39164924621582, b: 1.8858184814453125}
x: 0, pred: 1.8858184814453125
x: 1, pred: 7.277467727661133
x: 2, pred: 12.669116973876953
x: 3, pred: 18.060766220092773
示例 2:通过学习系数 a、b 和 c,使用 adadelta 优化器拟合二次方程。优化器配置如下:
- 学习率 = 0.01
- ρ = 0.2
- ε = 0.5
Javascript
// Importing tensorflow
import * as tf from "@tensorflow/tfjs"
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();
// Setting configurations for our optimizer
const learningRate = 0.01;
const rho = 0.2;
const epsilon = 0.5;
// Creating the optimizer
const optimizer = tf.train.adadelta(learningRate, rho, epsilon);
// 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: 3.1871466636657715, b: 1.5096971988677979, c:0.8317463397979736
x: 0, pred: 0.8317463397979736
x: 1, pred: 5.528590202331543
x: 2, pred: 16.599727630615234
x: 3, pred: 34.04515838623047
参考: https://js.tensorflow.org/api/1.0.0/#train.adadelta