Computer Science · A4.3 Machine learning approaches (HL only)

A4.3.3 — Hyperparameter Tuning and Evaluating Supervised Models

Unlock the full notes

Preview

IB DP Computer Science — A4 Machine learning — Machine learning approaches | HL | A4.3.3

📋 What You Will Learn

  • Distinguish model parameters (learned) from hyperparameters (set before training)
  • Explain the role of hyperparameter tuning in model performance
  • Construct and read a confusion matrix (TP, FP, FN, TN)
  • Calculate and interpret accuracy, precision, recall and F1 score
  • Explain overfitting and underfitting, and how tuning navigates between them
  • Explain why tuning must use a validation set, not the test set

🎛️ Why This Matters An oven "learns" nothing about your cake — but the temperature you set before baking decides whether it is raw or burnt. Hyperparameters are the machine-learning oven dials: they are not learned from the data, yet choosing them well or badly can matter as much as the algorithm itself.

Parameters vs Hyperparameters

  • Parameters are values the algorithm learns from the training data — e.g. the slope and intercept in linear regression, the split thresholds in a decision tree.
  • Hyperparameters are settings

Continue reading the full note

In this topic

More notes on A4.3 Machine learning approaches (HL only)

  1. 01A4.3.2 — Classification: K-Nearest Neighbours and Decision Trees
  2. 02A4.3.8 — Artificial Neural Networks: Perceptrons to Multi-Layer Networks
  3. 03A4.3.10 — Model Selection and Comparison in Machine Learning
  4. 04A4.3.7 — Genetic Algorithms: Evolution as an Optimiser
  5. 05A4.3.5 — Association Rule Learning: Uncovering Relationships in Data
  6. 06A4.3.9 — Convolutional Neural Networks: Learning Spatial Hierarchies
  7. 07A4.3.6 — Reinforcement Learning: Agents, Rewards and Policies
  8. 08A4.3.1 — Linear Regression: Predicting Continuous Outcomes
  9. 09A4.3.4 — Clustering: Grouping Data by Similarity