Learn AI Create the Future — with Pranay
Main site

All sections

The whole curriculum, grouped into 14 areas. Start at the top if you are new. If you came here for one thing, filter to its area or search — everything is on this page or one click from it.

Every lesson here is written by .

14 areas 117 sections 1,162 of 1,167 lessons published 10 new lessons every day

Showing Machine Learning — 11 sections, 105 lessons. Show all 14 areas

Machine Learning

How a machine learns from data, and the classical algorithms that still beat deep learning on most real tables.

11 sections 105 of 105 lessons published

Machine Learning

How computers learn patterns from data without being told the rules.

14 published

  1. What is machine learning?
  2. Supervised learning
  3. Unsupervised learning
  4. Linear regression
  5. Logistic regression
  6. Classification
  7. Clustering
  8. Decision trees
  9. Random forest
  10. XGBoost
  11. Overfitting and underfitting
  12. Train, test and validation splits
  13. Model evaluation
  14. Feature engineering

Classic Algorithms in Depth

The workhorse algorithms behind most real predictions, one lesson each, with the maths that makes them behave.

10 published

  1. K-nearest neighbours
  2. Distance metrics
  3. Support vector machines
  4. The kernel trick
  5. Naive Bayes
  6. Linear discriminant analysis
  7. Gaussian process regression
  8. How a tree chooses a split
  9. Pruning a decision tree
  10. Semi-supervised learning

Linear Models and Regularisation

Straight-line models are still the ones you ship when you need to explain the answer. Here is how to make them behave.

10 published

  1. Ridge regression
  2. Lasso regression
  3. Elastic net
  4. Multicollinearity and VIF
  5. Generalised linear models
  6. Poisson regression for counts
  7. Quantile regression
  8. Robust regression
  9. Regression diagnostics
  10. Bayesian linear regression

Ensembles and Gradient Boosting

Many weak models beat one strong model on tabular data. This section is why, and how to tune the ones that win competitions.

10 published

  1. Why ensembles work
  2. Bagging
  3. Out-of-bag evaluation
  4. AdaBoost
  5. LightGBM
  6. CatBoost
  7. Tuning gradient-boosted trees
  8. Monotonic constraints
  9. Feature importance done right
  10. Stacking

Preprocessing and Feature Selection

Everything that happens to your data between the CSV and the model, and how to pick the columns worth keeping.

9 published

  1. Feature scaling
  2. Power transforms
  3. Binning and discretisation
  4. The hashing trick
  5. Filter feature selection
  6. Recursive feature elimination
  7. Boruta and shadow features
  8. Sample weights
  9. Learning from data that does not fit in memory

Dimensionality Reduction

Squeezing hundreds of columns into a few, without losing the thing that mattered.

10 published

  1. The curse of dimensionality
  2. Principal component analysis
  3. Kernel PCA
  4. Truncated SVD and LSA
  5. Independent component analysis
  6. Non-negative matrix factorisation
  7. Random projections
  8. t-SNE
  9. UMAP
  10. Reading an embedding plot honestly

Clustering in Depth

Finding groups nobody labelled, and knowing how many groups there really are.

8 published

  1. K-means initialisation and local minima
  2. How many clusters?
  3. Gaussian mixture models
  4. Expectation-maximisation
  5. HDBSCAN
  6. Clustering mixed numeric and categorical data
  7. Comparing clusters to known labels
  8. Using clusters as features

Outlier and Anomaly Detection

Finding the handful of rows that do not belong, when almost nothing is labelled.

7 published

  1. Outliers vs novelties
  2. Z-scores, IQR fences and MAD
  3. Isolation forest
  4. Local outlier factor
  5. One-class SVM
  6. Anomaly detection by reconstruction error
  7. Evaluating an outlier detector

Calibration and Uncertainty

Getting a model to say how sure it is, and having that number mean something.

8 published

  1. ROC vs precision-recall curves
  2. Choosing a threshold from costs
  3. Reliability diagrams and calibration error
  4. Platt scaling
  5. Isotonic calibration
  6. Proper scoring rules
  7. Conformal prediction
  8. Prediction intervals for regression

Imbalanced, Multi-class and Multi-label

What changes when one class is rare, when there are twenty classes, or when a row can have several answers at once.

11 published

  1. SMOTE and its variants
  2. Undersampling strategies
  3. Class weights
  4. Resampling inside cross-validation
  5. Balanced bagging and EasyEnsemble
  6. Lift and gain charts
  7. One-vs-rest and one-vs-one
  8. Macro, micro and weighted averaging
  9. Multi-label classification
  10. Classifier chains
  11. Ordinal targets

Other ways to use this site

Learning paths

An ordered route through the lessons for one job: ML developer, AI engineer, computer vision engineer.

Projects

Build real things with full code, a dataset and a deployment guide.

AI glossary

Every term you keep seeing, defined in one plain sentence first.

Error database

Paste the error you got. Find out what it means and how to fix it.