Tensorflow.js tf.grads()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。它还可以帮助开发人员用 JavaScript 语言开发 ML 模型,并且可以直接在浏览器或 Node.js 中使用 ML。
tf.grads()函数接受一个函数f(x) 并返回一个函数gx
语法:
tf.grads (f)
参数:
- f:这是计算梯度的给定函数。
返回值:它返回一个数组。
示例 1:
Javascript
// Importing the @tensorflow/tjs library
const tf=require("@tensorflow/tfjs")
const f = (a, b) => b.add(a);
// Grad function is used
const g = tf.grads(f);
// Tensor is declared
const a = tf.tensor1d([5, 6]);
const b = tf.tensor1d([-10, -20]);
// Variables are defined
const [gfg1] = g([b, a]);
// Variable is printed
gfg1.print();
Javascript
// Importing the @tensorflow/tfjs library
const tf=require("@tensorflow/tfjs")
const f = (a) => a.mul(8);
// Grad function is used
const g = tf.grads(f);
// Tensor is declared
const a = tf.tensor1d([50, 60]);
// Variables are defined
const [gfg1] = g([a]);
// Variable is printed
gfg1.print();
输出:
Tensor
[1, 1]
示例 2:
Javascript
// Importing the @tensorflow/tfjs library
const tf=require("@tensorflow/tfjs")
const f = (a) => a.mul(8);
// Grad function is used
const g = tf.grads(f);
// Tensor is declared
const a = tf.tensor1d([50, 60]);
// Variables are defined
const [gfg1] = g([a]);
// Variable is printed
gfg1.print();
输出:
Tensor
[8, 8]
参考: https://js.tensorflow.org/api/latest/#grads