123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293
/** * This example trains a neural network to predict the output of the function y = 2x + 1 */
import {  Cost,  CPU,  DenseLayer,  Sequential,  setupBackend,  tensor1D,  tensor2D,} from "../mod.ts";
/** * The test data used for predicting the output of the function y = 2x + 1 */const testData = [20, 40, 43, 87, 43];// deno-lint-ignore no-explicit-anyfunction fmt(input: any) {  return (input.data as Float32Array).map((e: number) => Math.round(e))[0];}
/** * Setup the CPU backend. This backend is fast but doesn't work on the Edge. */await setupBackend(CPU);
/** * Creates a sequential neural network. */const network = new Sequential({  /**   * The number of minibatches is set to 4 and the output size is set to 1.   */  size: [4, 1],
  /**   * The silent option is set to true, which means that the network will not output any logs during trainin   */  silent: true,
  /**   * Creates two dense layers, with the first layer having 3 neurons and the second layer having 1 neuron.   */  layers: [DenseLayer({ size: [3] }), DenseLayer({ size: [1] })],
  /**   * The cost function used for training the network is the mean squared error (MSE).   */  cost: Cost.MSE,});
const start = performance.now();
/** * Train the network on the given data. */network.train(  [    {      // y = 2x + 1      inputs: tensor2D([[1], [2], [3], [4]]),      outputs: tensor2D([[3], [5], [7], [9]]),    },  ],  /**   * The number of iterations is set to 400.   */  400,  /**   * The number of batches is set to 1.   */  1,  /**   * The learning rate is set to 0.01.   */  0.01,);
console.log("training time", performance.now() - start, " milliseconds");console.log("y = 2x + 1");
for (const test of testData) {  /**   * Make a prediction on the test data.   */  const predicted = await network.predict(tensor1D([test]));  console.log(    `input: ${test}\noutput: ${fmt(predicted)}\nexpected: ${2 * test + 1}\n`,  );}