Introduction to Machine Learning

Lessons

1. 1Mathematical modeling SlidesDescribing real-world systems with equations so we can analyse, predict and optimise them.8 min
2. 2Example: the SIR epidemic model SlidesSplitting a population into Susceptible, Infected and Recovered and modelling the flow between them.10 min
3. 3Types of mathematical models SlidesDeterministic, stochastic, dynamic, optimisation, empirical, statistical and other families of models.10 min
4. 4AI vs. ML vs. Deep LearningAI is the broad goal, ML learns from data, and deep learning is ML with many-layered neural networks.7 min
5. 5Why machine learning?Problems too complex for hand-written rules, and patterns that change over time.6 min
6. 6Applications of MLVision, language, recommendation, finance, healthcare and science.7 min
7. 7Learning from dataUsing examples to fit a model that generalises to new, unseen cases.8 min
8. 8Datasets, features and labelsRows are examples, columns are features, and the label is what we want to predict.8 min
9. 9Parameters vs. hyperparametersParameters are learned during training; hyperparameters are chosen before it.7 min
10. 10Training, validation and test setsFit on training data, tune on validation data, report once on test data.9 min
11. 11Supervised, unsupervised and reinforcement learningLearning from labels, from structure alone, or from rewards.10 min
12. 12Classification vs. regressionPredicting a category versus predicting a continuous number.7 min
13. 13Parametric vs. nonparametric learningA fixed number of parameters versus models that grow with the data (e.g. k-NN).8 min

Key formulas

Key terms

Classification

Predicting a discrete category.

Deep learning

Machine learning with neural networks of many layers.

Feature

An input variable describing an example (a column in the dataset).

Hyperparameter

A setting chosen before training (learning rate, k, depth, λ), tuned on validation data.

Label

The target value the model should predict.

Parameter

A value learned during training, such as a weight or bias.

Regression

Predicting a continuous number.

Reinforcement learning

Learning a policy from rewards by acting in an environment.

Supervised learning

Learning to predict labels from labelled examples.

Test set

Held-out data used once to estimate final performance.

Unsupervised learning

Finding structure in data without labels.

Validation set

Data used to compare models and tune hyperparameters.