TensorFlow.js Playground

// Log TensorFlow.js version
(async function() {
  console.log("TF.js version:", tf.version.tfjs);
})();

// Creating and inspecting tensors
(async function() {
  const a = tf.tensor([1, 2, 3, 4], [2, 2]);
  const b = tf.scalar(42);
  const c = tf.tensor2d([[1, 2], [3, 4]]);
  console.log('Tensor a:');
  a.print();
  console.log('Scalar b:');
  b.print();
  console.log('Tensor c:');
  c.print();
})();

// Basic math ops on Tensors
(async function() {
  const x = tf.tensor2d([[1, 2], [3, 4]]);
  const y = tf.tensor2d([[5, 6], [7, 8]]);
  
  const sum = x.add(y);
  const product = x.matMul(y);
  
  console.log('x + y:');
  sum.print();
  console.log('x dot y:');
  product.print();
})();

// Basic linear regression with TF.js
(async function() {
  // We will learn y = 2x - 1

  // 1. Create a model
  const model = tf.sequential();
  model.add(tf.layers.dense({ units: 1, inputShape: [1] }));

  // 2. Compile the model
  model.compile({ loss: 'meanSquaredError', optimizer: 'sgd' });

  // 3. Generate data
  const xs = tf.tensor2d([1, 2, 3, 4], [4, 1]);
  const ys = tf.tensor2d([1, 3, 5, 7], [4, 1]);

  // 4. Train the model
  console.log('Training the model for 10 epochs');
  await model.fit(xs, ys, { epochs: 10 });

  // 5. Predict
  const output = model.predict(tf.tensor2d([5], [1, 1]));
  output.print();
})();

// Simple Multi-layer Network for XOR
(async function() {
  
  // XOR dataset
  const xs = tf.tensor2d([
    [0, 0],
    [0, 1],
    [1, 0],
    [1, 1],
  ]);
  const ys = tf.tensor2d([
    [0],
    [1],
    [1],
    [0],
  ]);

  // Create a simple multi-layer model
  const model = tf.sequential();
  model.add(tf.layers.dense({ units: 4, inputShape: [2], activation: 'relu' }));
  model.add(tf.layers.dense({ units: 1, activation: 'sigmoid' }));
  
  model.compile({
    loss: 'binaryCrossentropy',
    optimizer: 'adam',
    metrics: ['accuracy'],
  });
  
  // Train
  console.log('Training the model for 100 epochs');
  await model.fit(xs, ys, {
    epochs: 100,
    verbose: 0
  });
  
  // Evaluate
  const evalOutput = model.predict(xs);
  console.log('Predictions:');
  evalOutput.print();
  
  // Round to [0 or 1]
  const rounded = evalOutput.round();
  console.log('Rounded predictions:');
  rounded.print();
})();