A4 · machine learning

Machine learning you can try for yourself

Each demo lets you try out one algorithm or idea from the A4 syllabus. Change the settings and watch how the algorithm responds.

• A4.3.2 HL

  k-nearest neighbours

  Add labelled points and change k. Watch how the class with the most votes among nearby points determines each decision region.

  Open→
• A4.3.1 HL

  Linear regression and gradient descent

  Watch gradient descent adjust a straight line to fit the data. Change the learning rate and see how the error changes.

  Open→
• A4.3.4 HL

  k-means clustering

  Watch k-means assign points to clusters and move each centroid to the mean of its assigned points.

  Open→
• A4.3.8 HL

  The perceptron

  Train a single perceptron on a logic gate and watch its boundary line rotate into place. Then see why XOR needs a hidden layer.

  Open→
• A4.3.7 HL

  Genetic algorithm

  Use selection, crossover and mutation to evolve random text towards your target. The best candidates are kept unchanged, so the best fitness can improve or stay the same.

  Open→
• A4.3.6 HL

  Reinforcement learning

  An agent learns a route across a grid by trial and error, using rewards and penalties. You set the balance between exploration and exploitation and see the move it learns for each square.

  Open→
• A4.3.9 HL

  Convolution and feature maps

  Watch a convolutional filter build a feature map, then use max pooling to reduce its size. These operations help CNNs process image features.

  Open→
• A4.3.3 HL

  Overfitting and evaluation metrics

  Change the degree of a polynomial from underfitting to overfitting and compare training and test error. Then move one threshold and see accuracy, precision, recall and F1 score change.

  Open→
• A4.4.1

  Accuracy versus fairness

  Train a classifier on a biased loan dataset, compare approval rates and agreement with past decisions, and investigate why matching history does not prove fairness.

  Open→