ML Foundations: 20 MCQs with answers

A way for computers to learn from data without being explicitly programmed

Machine learning is about teaching computers to learn patterns from data, not just following fixed rules.

Teaching a computer to recognize cats using labeled photos

Supervised learning uses labeled data, like photos with 'cat' or 'not cat' labels, to train a model.

The answer we want the computer to predict

A label is the correct output for a given input, used in supervised learning to teach the model.

An individual measurable property of the data

Features are the inputs used by the model, like height and weight for predicting health.

Feeding data to the algorithm so it can learn patterns

Training is the process where the algorithm learns from data by adjusting its parameters to reduce errors.

Supervised learning

Supervised, unsupervised, and reinforcement learning are the three main categories; the others are not standard types.

To make accurate predictions or decisions from data

The goal is to learn patterns from data to make predictions or decisions without explicit programming for each case.

Spam detection in emails

Spam detection uses ML to classify emails based on learned patterns, unlike rule-based software.

The learned pattern or function that maps inputs to outputs

A model is the result of training—a mathematical representation that makes predictions from inputs.

The proportion of correct predictions out of total predictions

Accuracy is a simple metric: correct predictions divided by total predictions, often used for balanced datasets.

False

Traditional programming uses explicit rules; ML learns from data to create its own rules.

True

Supervised learning uses labeled examples to teach the model the correct output for each input.

Supervised uses labeled data, unsupervised uses unlabeled data

The key difference is the presence of labels: supervised learning has labeled examples, unsupervised finds patterns without labels.

Clustering customers into segments based on behavior

Clustering is a classic unsupervised task because it groups data without predefined labels.

To evaluate how well the model performs on unseen data

A test set is held out from training to check generalization—how the model does on data it hasn't seen.

When the model learns the training data too well, including noise, and fails on new data

Overfitting means the model memorizes the training data rather than learning general patterns, hurting performance on new data.

A reward or penalty

Reinforcement learning uses rewards (positive) and penalties (negative) to guide the agent's behavior.

A subfield of machine learning using neural networks with many layers

Deep learning uses deep neural networks with multiple layers to learn complex patterns, a subset of ML.

To evaluate the model's performance on unseen data

Holding out a test set lets us see how well the model generalizes to new, unseen examples.

Training set is used to learn parameters; validation set is used to tune hyperparameters

The training set updates the model's weights, while the validation set helps choose hyperparameters and avoid overfitting.