Home / Summaries / COMP 3711 / classification-labels-training

Classification - General Framework

Q: Define classification as a machine learning task. What does the model learn?

Classification is the task of learning a target function f (a classification model) that maps an input attribute set x to one of several predefined, discrete class labels y. The model learns the relationship between input features and class membership from labelled training examples, then applies that relationship to predict labels for unseen records.

4 important questions on Classification - General Framework

Define classification as a machine learning task. What does the model learn?

Classification is the task of learning a target function f (a classification model) that maps an input attribute set x to one of several predefined, discrete class labels y.
The model learns the relationship between input features and class membership from labelled training examples, then applies that relationship to predict labels for unseen records.

What are the two purposes a classification model can serve?

Descriptive modelling — explains what features distinguish objects of different classes (e.g., what attributes separate mammals from reptiles). Used for understanding.
Predictive modelling — predicts the class label of previously unseen records. Used for decision-making.

What are the two phases of classification, and what happens in each?

Induction — a learning algorithm is applied to a training set (records with known labels) to build the model
Deduction — the learned model is applied to a test set (records with unknown labels) to predict their classes

Think: induction = building the model; deduction = using it.

What key assumption underlies model evaluation?

The stationarity assumption: future (test) examples are drawn from the same distribution as the training data. If the test distribution differs, evaluation metrics are misleading — the model may perform well on the test set but fail in deployment.

The question on the page originate from the summary of the following study material:

COMP 3711

View summary