Learn ML / Stage 3
3 · Machine learning
Classical machine learning and the statistical view of learning from data. Builds on stages 1 and 2.
By the end of this stage, you can…train and evaluate a regression and a classification model with scikit-learn, choose a sensible metric and validation split, explain overfitting and the bias-variance trade-off, and say when a simpler model is the better choice.
What you will learn
- Supervised learning Essential
- Unsupervised learning Essential
- Model evaluation Essential
- ML in practice Essential
- Probabilistic and Bayesian ML If you need it
- Causal inference If you need it
- Reinforcement learning If you need it
- Data science If you need it
Resources per subject
Supervised learning Essential
- Yaser Abu-Mostafa, Learning from Data, CaltechA complete Caltech course on what learning is, why it works and how to do it well: from the perceptron and linear models to overfitting, regularisation and validation.★ recommended · videos · notes · exercises
- Kilian Weinberger, Cornell CS4780: Machine Learning for Intelligent Systemsvideos
- Cornell CS4780 lecture notesnotes
4 more options
- scikit-learn user guidedocs · examples
- An Introduction to Statistical Learning (James, Witten, Hastie & Tibshirani)free book · Python labs · free video course
- Stanford CS229: Machine Learningnotes · videos
- MIT 6.036: Introduction to Machine Learningvideos · notes · exercises
Unsupervised learning Essential
- Stanford CS229 lecture notes: unsupervised learningk-means, mixture models and EM, PCA and ICA, with the maths.★ recommended · notes
- scikit-learn: unsupervised learningdocs · examples
Model evaluation Essential
- Sebastian Raschka, Model Evaluation, Model Selection, and Algorithm Selection in Machine LearningA thorough review of validation splits, cross-validation, the bootstrap and statistical tests for comparing models.★ recommended · free paper
- scikit-learn: model selection and evaluationdocs · examples
- Machine Learning Crash Course, Googlecourse · exercises
ML in practice Essential
- Christian Kästner, Machine Learning in Production, Carnegie Mellon UniversityA Carnegie Mellon course on taking models from notebook to product: testing, deployment and monitoring.★ recommended · course · notes
- Full Stack Deep Learningvideos · deploying models
- Made With MLcourse · MLOps
1 more option
- Kaggle Learnshort interactive courses
Probabilistic and Bayesian ML If you need it
- Philipp Hennig, Probabilistic Machine Learning, University of TübingenA complete lecture series on Bayesian inference, Gaussian processes and probabilistic models for ML.★ recommended · videos
- Statistical Rethinking (McElreath)videos · course materials
- Probabilistic Machine Learning (Murphy)free books
3 more options
Causal inference If you need it
- Introduction to Causal Inference, Brady Neal (Mila)Written for people with an ML background, with videos and a free book.★ recommended · videos · free book
- Causal Inference: What If (Hernán & Robins)free book
Reinforcement learning If you need it
- David Silver, UCL / DeepMind, Reinforcement Learning courseThe classic university course. Use Sutton & Barto alongside it, then the 2021 DeepMind series for newer methods.★ recommended · videos · slides
- Reinforcement Learning: An Introduction (Sutton & Barto)free book
- DeepMind x UCL Reinforcement Learning Lecture Series 2021 (van Hasselt, Borsa, Hessel)13 video lectures
3 more options
- CS285: Deep Reinforcement Learning, Berkeleyvideos · slides · homework
- Spinning Up in Deep RL, OpenAIdocs · code
- Deep RL course, Hugging Facecourse · code
Data science If you need it
- Introduction to Data Science in Python, University of Michigan (Coursera)Practical data handling with pandas.★ recommended · videos · exercises · free to audit
- Data Science Specialization, Johns Hopkins University (Coursera)videos · exercises · free to audit
- The Data Scientist's Toolbox, Johns Hopkins University (Coursera)videos · exercises · free to audit
3 more options
- R Programming, Johns Hopkins University (Coursera)videos · exercises · free to audit
- Daniel Lakens, Improving Your Statistical Inferences, TU Eindhoven (Coursera)videos · exercises · free to audit
- Data Science, IBM Cognitive Classvideos · exercises