Skip to main content

Topic roadmap

Machine learning

15 lessons, about 63 hours in all, from the basics to advanced. Start with the first, or jump to what you need.

Lessons, easiest first

  1. Machine learning · Beginner 3 · about 6 hMaths for machine learningBuild intuition for the three areas of maths behind machine learning (vectors and matrices, derivatives and gradients, and probability) well enough to understand how models learn and to read ML papers and docs.
  2. Machine learning · Beginner 3 · about 5 hStatistics for machine learningUse the statistics ML engineers need daily: describing data with the right averages and spreads, sampling, correlation versus causation, confidence intervals and A/B tests.
  3. Machine learning · Intermediate 1 · about 3 hChoosing a metricChoose evaluation metrics that match the business problem: accuracy, precision, recall, F1, ROC and PR AUC for classification; MAE, RMSE and MAPE for regression; ranking metrics; and cost-weighted measures.
  4. Machine learning · Intermediate 1 · about 4 hClassificationBuild classifiers that predict categories, from logistic regression to trees; choose a decision threshold from business costs; and handle imbalanced classes such as fraud.
  5. Machine learning · Intermediate 1 · about 3 hCross-validation and leakageEstimate model performance honestly: train, validation and test splits, k-fold and time-based cross-validation, grouped splits, and how to find and prevent data leakage with pipelines.
  6. Machine learning · Intermediate 1 · about 5 hFeature engineeringTurn raw records into features that help models: aggregates over time, ratios, date parts, encodings for categories, text and location features, while avoiding leakage and keeping training and serving consistent.
  7. Machine learning · Intermediate 1 · about 4 hRegressionBuild and judge regression models that predict numbers: linear regression, the meaning of coefficients, error metrics such as MAE and RMSE, and the traps of outliers, non-linearity and extrapolation.
  8. Machine learning · Intermediate 1 · about 5 hTrees and gradient boostingUse decision trees, random forests and gradient-boosted trees (XGBoost, LightGBM) for tabular data: how they work, how to tune them, and how to explain their predictions.
  9. Machine learning · Intermediate 1 · about 4 hUnsupervised learningFind structure in data without labels: cluster customers or products with k-means and other methods, reduce many features to a few with PCA, and detect unusual records, then judge whether the result is useful.
  10. Machine learning · Intermediate 2 · about 5 hCNNs for imagesUse convolutional neural networks for image tasks: how convolutions detect patterns, why transfer learning from pre-trained models is the default, data augmentation, and evaluating image models properly.
  11. Machine learning · Intermediate 2 · about 5 hNeural networksExplain how a neural network turns inputs into predictions through layers of weights and activations, how it learns with a loss, backpropagation and gradient descent, and how to spot overfitting.
  12. Machine learning · Intermediate 2 · about 5 hPyTorchTrain and save a model in PyTorch: tensors and devices, datasets and data loaders, defining a model, the training loop, evaluation mode and checkpoints.
  13. Machine learning · Intermediate 2 · about 4 hTransformersExplain how transformers work (tokens, embeddings, attention, positional information, stacked layers) and how the same architecture powers language models, embedding models and vision models.
  14. Machine learning · Pro 1 · about 2.5 hModel size and scaling laws: parameters, training data and computeRead a model's size, estimate the memory and training compute it needs, explain the Chinchilla rule and why models are now trained far past it, and judge when a smaller model is the better choice.
  15. Machine learning · Pro 1 · about 2.5 hPerplexity and cross-entropy: how language models are measuredExplain cross-entropy and perplexity in plain words, compute perplexity from token probabilities, compare models fairly with bits per byte, and use these numbers to pick base models, filter data and spot contamination.

On these role roadmaps

Interview questions on machine learning

Books on machine learning

All recommended books

Today's AI turns are used

Lessons, roadmaps, videos, practice cards and the code playground stay free with your account, with no limit.